The Nursing Process

>> patricia grady: okay, now recall the --there's been a great deal of interest around the counsel table and across the instituteabout some of the things that are happening in big data. and we've been spending a great deal of timeand committees on campus -- and for several years trying to figureout how to anticipate this, what to do, how do we prepare, et cetera. and so last time at councilyou heard from eric greene, who was the acting chief information officerand heard some of the things that we're doing on campus.

you also heard about some of the work groupsthat have been called in and some of our representatives. and so we're very lucky today. we feel very fortunate thatwe have dr. patty brennan, who has represented us insome of the trans-nih activity, to come to talk to us about big data innursing science and to describe some of her experiences and some of her thoughts onthe idea and to help to point the way for us about directions that we might go andways that we might capture this opportunity. most of you know patty brennan fairly well.

you may or may not know thatshe actually has her ph.d. in industrial engineering, which mostyou probably think she's just good at this because she's learned italong the way, but actually she, in addition to her nursing background,has the engineering background as well. and she's really well-known for so many ofher activities and some of the pioneering things that she has done. many of those involve initials and acronymsso i won't go through them because she'll tell us a little bit about that. but i do want to say that shedeveloped the computer link,

which is an electronic network designed toreduce isolation and improve self-care among home-care patients, andalso directed heartcare, which is a web-based tailored informationand communication service that helped home-dwelling cardiac patients recover fasterwith fewer complications and fewer side effects. those are just a few ofher many accomplishments, but those are particularlynoteworthy and in tune with our mission, so patty, we're looking forward tohearing what you have to say today. thanks so much.

>> patricia brennan: thank you very much. i appreciate the opportunity to give backto ninr what they've been giving me for 30 years. since -- in 1982 i received a doctoraltraining grant to study the beginning what was at the time we thoughtdecision support systems, which we thought weregoing to be here any day. and i hear they're going to be here any day. [laughter] i'm here again to tellyou they're going to be here any day. the computer link project that pat mentionedis a project that we started in 1987 to use,

at the time, terriblyhorrible computer systems, wise-30 terminals and 110 bod modems tosupport people living with aids at a time that there was a desperate need to extendnursing into the lives of people and no real way to do that. and not to do it by what theprofessionals believed was important, but to do it by what the patients needed. so we received support for six months ofstudy in and out of the daily lives of people living with aids. it helped us form the foundationof an exceptional computer network.

and ncnr supported that also. why i told you this little story of myhistory is to tell you that ninr and its predecessors have been onthe cusp of information, informatics extending into nursing andinformatics-enabled nursing practice for a long time. this is not new. there's just some new terms, some new ideas,but fundamentally we've been there and we need to stay there. we need to hang on to the table and we needto have the place there to bring what we know

as well as to learn what was presented to us. i'm going to spend about 20 minutes talkingabout some of the experiences with the bd2k initiative in the last couple of months hereand also to talk a little bit about where i'd like to see nursing science go. and then we have time for you to talk and ireally do want to hear what we need to do in publications, publicpolicy, in nursing research, in training ourundergraduates around this initiative. i do not expect that anyone in this roomhas any more answers to this project and idea than i do, so we're all even,the playing field is flat here.

but we all come with eithera desire for, a belief in, or an experience as nurses and that meanswe're focused on two things: we focus first and foremost on thehuman response to disease, illness, injury, developmentalchallenges, family structures, community issues. we focus on something that nobody else does. and there's big data thathelps us understand that. secondly that we bring a cultureand it is a culture of engagement, of respect, of empowerment that doesnot exist in other areas where the bd2k

conversations are going on. so we must be present. it is responsible as our profession to be inthe conversations to ensure that what we know continues to inform the big data initiative. to let you know, no surprise to anyone, thatthe amount of data in the world doubles every 18 months to two years. doubles. now, it used to be thought, well let's --everyone likes to see things doubling quickly and when i first came out as apracticing nurse in critical care in 1975,

people said then that the half-life ofcritical care knowledge was 14 months, which meant after about two and a halfyears you knew nothing to take care of those patients. that was a bad thing. we needed technologiesto come in to support it. but now what's happening is the data isflowing at us much more rapidly and in much more diverse and disparate manner than itever has before and we don't have the funnels to push that datathrough to make sense of it. we don't have in place themodels to understand it.

and we don't have thecomputational power to manage it. that's what we're here totalk a little bit about. believe the bd to care, the big datainitiative probably is about 12 or 13 years old now in the u.s. in different sectors. and it's driven by a beliefthat there's a pony in there. if we just got that dataand looked at it enough, we'd find something, truth,knowledge, money, god knows what, but there's a belief thatthere's a pony in there.

hang onto to your slightlevel of skepticism about that. i hope there's a pony in there. but i'm not sure. and we need to be careful as we look at ourprecious nursing resources and the questions we try to answer, that they pursue questionsto address the phenomenon of nursing rather than pursue participation in activities. now, walmart does onemillion transactions an hour. so since we've been sitting here walmarthas done almost 1.5 million transactions. facebook has 50 billionpictures that they deal with.

there's data. there's all the different kinds of data. and that data tells usstories of people's lives, gives us indicators of howthey're managing in the world, and could be potentiallyvaluable to us and our patients. we need to think abouthow to get a hold of that. now, in nih's approach to thebig data to knowledge initiative, they've made some constraints about what'sdefined as big data and i think these are hypothesized or conjectures asstarting points rather than ending points.

so i'm going to take you from the ideathat the amount of data in the world, that is pings coming out ofevery cell phone in this room, numbers of beeps of an electronicmonitor on someone's cardiovascular system, are coming -- doubling faster and faster. and at the same time, what nih has said is wethink this is important for the purposes of knowledge and knowledge-building. so in the nih bd2k model the idea isthat big data comes from three key sources. first of all, that's there's projectsthat are intended specifically to produce important resources forthe research community.

this is creating theselarge data sets, data centers, coordinating centers thatshould and could be reused. and they're important for us tounderstand and to know about. and in addition, large data sets are createdin individual projects but they might be useful more broadly if therewere standard nomenclatures, terminologies, models tohelp them be understood. so one -- the first bullet addressesthe explicit development of data sets. the second bullet looks at the deriveddata sets from individual research projects. and the third data set is probablywhat we're most familiar with in nursing,

small data sets that start off small butcan grow quickly when they're amplified or aggregated or integratedwith other data resources. so big data from the nih perspective is datathat is deliberately sought for the purposes of knowledge-building. and it's sought either in deliberateinvestments for creating data repositories or in small projects that areunderstood in a way that can be shared. easier said than done. what makes data big? well, first of allthere's often thought of this,

is there's just uncountable elements, there'sthousands and thousands and thousands of elements. my current research, which was funded on afriday afternoon in september right before everything else shut down -- my current dataproject studies how -- looks at how we can look at home environments by using a speciallaser technology to capture three-dimensional images of the home and recreatethem in a virtual reality cave. we collect 95 million datapoints from every household we go to. you can't run that on atablet, i can tell you right now. it's a lot of data.

i'm not sure all of it makessense and all of it's useful, but there are uncountable elements in there. more importantly though big data often has ahigh variety of data types and data sources, that it is not just having alot of the same kind of data, but it's having a lot of different kindsof data that comes from a lot of different places. so issues such as provenance --where did this data come from, how do we trust it, how do we knowwhose it is -- become extremely important. for theinformatics-thinking people in the room,

these are not new questions. for the larger community theyactually are new questions. dan macey and his leadership on one ofthe bd2k work groups identified bd2k -- i'm sorry, identified big data as reallyhaving two fundamental characteristics. first of all, it exceeds the capacity ofunaided human condition for comprehension. that is, we can't think about it on our own;we can't keep the whole data set in our mind. most of you are like me, had a data modelthat you could actually think of all the pieces in your mind, if youcould think of every element. you could keep it in your head.

that's changing. you can't do it anymore. one brain isn't enough. secondly, that big data strains currentcomputational technologies and if therefore bound by the amount ofthe processing units, cpu, the bandwidth, the pipes thatwe can push the data through, or our storage ability. now when the hadron collider project cameto its active stage about two years ago, we began to realize the data that's createdin that collider can never be collected and

stored. there is too much data coming out. and that's actually going to be comingtrue in a number of areas as we begin to have ubiquitous sensing, as we begin to reallywant to understand not only how does a person walk, but how many times dotheir flutter, does their heartbeat, as they are walking. we add so much data to the moment that ourideas of let's store it in case we need it have got to go away. it's not going to work anymore.

there's not enough space for it. but more important than volume is the -- andthe diversity of data is really influential. it's really different when we talk aboutwhat does engagement or temperature mean? when you're in a home care situationtemperaturing my kid's head for it -- my kid's forehead feels hot. when you're in the critical caresituation maybe it's an inner ear probe. there's a different type of data that'scoming to the same point in a computation. but most important is what arethe questions we're going after? what is it we want to know from data?

now, understanding how to link big dataand the questions relevant to it require a concurrent and engaged interaction betweenthe question asker and the data resources. i'm going to ask you to keep both in mindtoday because what's critical for us in our discipline and in our roles as investigatorsand scientists and nursing is that we understand that the data exists for a purposeand it is our professional knowledge and our judgment that helps to create the questionsthat leads to answer from that data set. analyzing data is not enough. serendipitous discoveryof patterns, important, they're spotlights that mightgive us a clue to what to go after,

but fundamentally as scientists we need to beasking questions not simply swimming in data streams. now, the bd2k initiative at nih ispurported to enable biomedical scientists, including us, so i'm surethey mean medicine broadly, to capitalize more fully on the bigdata being generated by those communities, but often those research communities arenot research communities that are familiar to nursing. they are more often research communitiesthat are driven by genomic science, to some extent by environmentalscientists, the fundamental physiologies,

and while we have nurses who work inthose areas and nurses at the intersection, fundamentally the big data that ourdiscipline deals with doesn't appear on the roadmap yet for the big data initiative here. we're going to come back tothat before i close my remarks. but speaking of roadmaps: if you, like me,had trouble figuring out exactly do people mean when they say big data, youmight be relieved to see this diagram. this diagram -- i can send the link to if youwant to read it more clearly -- this diagram is the product of a gentlemannamed swami shandra carson. i said i wasn't going to try to say his name,it's so hard to pronounce and i apologize for

mutilating it. he basically took the idea of a metro map inwhich -- those of you who know i just came back from paris -- thismakes perfect sense to me now. the metro map tells us how to gothrough the big data experience. from the lower left-hand side, fundamentalsand statistics leading to machine learning -- that bright green line --up to text mining and nlp. this is a set of analytical tools. on the right-hand side, upper right handquadrant we have data management and data integration strategies, which includeeverything from how do we sample from a large

data set to how do we use tools likeprinciple components analysis to break things apart. the orange pathway isthe visualization pathway. the brown pathway is the set of tools andtoolboxes that are used to assist with big data assessment. and finally at the lowerbottom part here, the bright pink, slightly in the lower right-handcorner, we see the big data tool sets. these are -- sorry, thebig data storage sets. these are thecommunity-driven repository models,

data models that allow big data to beintegrated since the big data systems largely exist across a number of clinicaland computerized information systems. what you're seeing here is what swami arguesis the roadmap to coming a data scientist, the things that one needsto know as a data scientist. and fortunately someone has taken this metromap and attached your [unintelligible] to every stop. so, again, i can provide you with access tothe site that describes everything from the [unintelligible] computational tool setto where you can get a really good natural language processing parser.

what we're looking at here isa different view on science. notice that there's noquestion askers in here, there's question answering tools. notice that they're -- the ideas ofdefinition come more as an extraction, the natural language processing and textmining strategies rather than from the formalized models andframeworks that we use in nursing. the game is changing outside of us. and what part of our job is to figure out howwe want to either make sure the game stays broad enough that knowledge discovery remainsdriven by models and frameworks and theories,

which is our strategiesmore familiar in our field, or that we inform this level of computationalapproach to data by also bringing in the frameworks and understand andphilosophies that we have in nursing. now, for this reason and this reason alone,every nurse in the country who's a solid researcher could participate inany one of the bd2k initiatives. for reasons not clearly known to me,linda hardy contacted me and said, "can you do this? can you go to some of these meetings for us? we had some good interactions."

and i went, "i am not a data scientist." so i'm going to bringing to you nowthe perspective of a patty brennan, who's a solid scientist, and a goodthinker, but will tell you what some of the experiences that i had in the bd2k. and the first is the bd2k initiatives that --conversations that i've been in largely are at this level, and notat one of the questions. and why do the questions matter andhow do we get to the right questions, and that is why we need to be there. i kept sticking my hand up.

"what are the questions? why are we asking this? what are we doing with this?" and they -- they're -- theconversations can move and they do move. there's excellent people involved in this. so there were three workshops thatwere held at nih since last summer. the workshop on enhancingtraining for biomedical big data, the workshop on enablingclinical research -- sorry, research uses of clinical data, and thenthe frameworks for community-based standards

efforts. these are only one part ofthe larger bd2k initiative. you've been aware of some of thecenters that are being developed, the computationalservices, the training services. i'm going to focus on myexperience in these three. these are the ones that i know the best. so first, the workshop on enhancingtraining for the biomedical big data. this group has actually gone so far as toproducing their report and they have a really good set of resources online.

so they've been really quite helpful. and i'm really pleased to say with theboot camp planned this summer and some of the other conversations i've been in, i thinkninr is actually well-aligned with some of this, but there's a couple ofthings that are troublesome to me. so let's first of all talkabout what do trainees need? what came out of this meeting is thatfundamentally the trainees need to understand the processes andframeworks for data integration. often we think about data integration fromthe standpoint of having an index variable, usually a patient name or a casenumber, and then we integrate across that.

but there are lots of other ways to integratedata at the level of the individual, at the level of the phenomenon, and musen andhis group at stanford have done the most in the country to advancing the ideasof ontology and how those ontologies, those ways of thinking about phenomenon, canbe brought to a practical level through data integration. trainees need to understand how to modelbiomedical domain and helping students and train -- and our doctoral students understandthe difference between knowing a domain and modeling a domain isactually quite challenging. when you try to help people model a domainyou identify the key features in the domain

and the relationshipbetween the key features. sounds a lot like nursing theory, doesn't it? but when we talk about computational modelingand how we create computational models for domains, it requires an additional level ofthinking about the behaviors that we see, the phenomena, are they stable orunstable, the extent of stability, what do we do with uncertainty,what do we do with unpredictability? the group also argued that trainees need tounderstand standard metadata descriptions and standard ontologies. now, we know a lot about metadata right nowfrom the nsa and we're talking actually about

the exact same thing. that is, how do we knowwhat the data are about? how do we describe what the data areabout, the tags that go along with the data elements, understanding not what thedata are but what -- this is a body part, this is an experience, thisis a patient-reported outcome. that is really more what metadata refers to. ontologies, as i've said, they'restructures, they're ways of organizing data. some nurses have done -- begun to do work inontological development particularly around the computation ontologies.

when we say ontology from acomputer science standpoint, it draws from but is slightly differentfrom the philosophical concept of ontology. ontology in a computational standpoint bringswith it a structure that describes property such as inheritance. if i know this about you, cani also know that about you? can i derive other thingsabout this data element? it's important also to understand -- tobe able to evaluate the appropriateness of existing ontologies for thedata integration activities. one of my students, jane pisu, wastrained also at unc and nir post-doc,

studied the application ofontologies to family history. and she found that many of the ontologyproblems -- ontologies as existed had problems because they didn't allow for threekey things that nurses are well aware of. first of all, that whenyou get an organ transplant, you get genetic material with it. we don't quite know how to representorgan transplant on ontologies of family. secondly, that some familiesaren't what they think they are, that your sibling may actually not be thechild of both of your parents but may be the child of your parent andsomeone else -- 10 percent,

they suggest. and third and most important,that the ontology -- sorry, that the family structures that are importantmay be less the biological heritable ones and more the environmentones: who did you live with, did they smoke, did they go outto dinner on saturday nights, did they have activities on sunday? so the ontologies needed to bebroader and jane's work was great for that. the initiatives around the i2b2 program, theinformatics for the integrating of biology and bedside, have been particularly helpfulin helping us bring together models of

biologicals as well as genetic aspectsof an individual and models of care. however, they tend to still focus on clinicaltherapeutics that are more common outside of our discipline, lessrelated to our discipline. oops, i've shut myself off. now, the training workshop also recommended-- and this is going to be a challenge for us frankly in nursing -- that we havemultidisciplinary training approaches at the intersection of fieldsrather than discipline-specific. developing skills to workeffectively in team science, this has been very important.

[unintelligible] has been really stressingthis in the t-32s and we need to consider -- continue that. dual mentoring, that is havingan analytical or a data mentor, as well as a domain like a nursingmentor, is something that was encouraged, and also, creating a very diverse workforce. not everybody needs to know everything,but everybody needs to know who knows what things. i'm going to -- i'm not going toread this through extensively to you, but i just want to bring threepoints up to the training content.

there's no one way toimplement big data training. there's no one way to do this. there's no right -- one single right way. there's lots of good ways. but there is a need tostress fundamental research, which we already do in our programs,ensure access to large data sets which are of multiple types which is a littlebit more challenging for us to do. but the final one actually worried me. desire to add such trainingto medical school curriculum,

and this is a direct quotefrom the report by the way, that this might be easier to add suchtraining to the pre-medical experience. we need to think about how to broaden thosewords because if we're only going to think of this as a physician-scientist strategy,or if the bd2k initiative restricts it, we could lose a lot of knowledge that couldbe very important to understand the human response. long term training. the bd2k training group actually stressedtraining as a beginning of a lifetime learning process so that we build infundamental skills and periodically do

different things acrossthe career trajectory. four critical aspects here: organizing teamswith complementary experience so that our ph.d. and post-doc trainees get to understand whatit's like to be on a team that's working with big data, rather than toget the skills themselves. boot camps. now, they use the idea of bootcamps in a different manner. they suggested boot camps should be held atthe beginning of training programs to help, if you will, cross-fertilize the teamswith different kinds of backgrounds that are

needed. bringing groups together to solve difficultdata analysis problems and rather than to prove that they know how to doit, to determine a way to do it, is a critical strategy. and finally, to -- this is -- comesfrom a biomedical training model, but it actually fits forall of our nursing education. have students pursue rotations invarious laboratories or various groups. we often use a very tightly-held apprenticemodel in training nurse researchers. and in fact it's time to break that out alittle bit and make sure our students move in

and out of other groups. short term training. this is in the handout that you got. different aspects are considered important. but i want to call yourattention to the bottom line, code-a-thon-like intensive experience. weekend hack-a-thons,weekend immersion experiences, 10-hour manages, where people get togetherand push through solving one problem. this is a differentapproach than we use in nursing.

it might be actually useful in a number ofareas to think about how we could use an intensive mixed group tocreate not only an answer, but people with the skillsto come up with new answers. to summary training then, it shouldbe interdisciplinary and team-based, focused on the diversity, use alreadyopenly-accessible data and consider flexibility inapproaches to multiple training. these -- it doesn't soundlike rocket science does it? it's really there's notmuch magic bullet in here, accept that all of it revolves around now, ascale and an order of magnitude much greater

than we're used to working with. the other two workshops that were heldhave not produced their final reports. so i'm going to give you ashorter report about them. the -- i was very excited to be a part ofthe enabling research use of clinical data workshop. i thought this was the best place for nursingin general and for my group in particular to have a say in shaping the future. i have to say i came away with very mixedfeelings about how well we were able to do that.

first, let me tell you aboutthe discussions during the day. this is a team of people who met for a dayand a half and they talked about health from the perspective of information. that is, we can learn about the health ofpeople by studying the information we have about them. that premise is questionable. a person is not an emr. if we're going to derive health by simplylooking at the emr or looking at recorded data sets from individuals, we will not getthe picture of people that nurses are used to

seeing, first of all,because the emrs are incomplete, and secondly, because even if they werecomplete they only deal with a slice of a person's life. most of the time peopleare not in the hospital. so we need to figure outdifferent ways to add to this. i have to say the -- this particular bd2kgroup focused very specifically on the use of electronic health records and veryspecifically on the use of electronic health records in acute care environments. so the starting point is a little more narrowthan we're going to need and we need to help

them push that out. secondly, there was a strong believe thatthe electronic health record is complete, accurate, and useful. this is not always the case. maybe here. but largely the beliefthat ehrs are complete, accurate, and truthful actually comesfrom the idea that if you get enough of them together you'll -- the errors inthe variability will even out. well, nursing works atthe uniqueness of people,

right? we do -- we kind of work onthe fringe, not in the center. so we have to think about firstof all, how do we create a record, an accessible record of a person that bringstogether information about that fact that my cell phone's been pinging, that yesterday iwas in paris and today i'm in washington, tomorrow i'll be in madison, as well as thefact that three weeks ago i did go to see my nurse practitioner for my annual physical andthere is a legal electronic record for that. how do we bring that together? we do -- there is a great interest in -- fromthis group present to get at the phenotype

expressions of people and actually most ofwhat was being discussed are things that we actually deal with, thephenomenon of concern of nursing. we deal with these phenotype expressions, andso we have to help to bring that conversation broader. finally, there is a great interest insomething referred to as practical clinical trials. and i tried to discern exactly what thepeople mean by practical clinical trials. and largely what this means is a secondaryanalysis of clinical records to determine if a problem or response to a medicationor expression of a particular disease,

could have been detected earlier had webeen looking at or minding this data. and the one that's most commonly used -- heldup -- and there's many really good examples of this now -- but the cardiacproblems following celebrex. if you really were looking at the clinicaldata well enough you actually could have detected the cardiac problems were going tocome and when they were even going to begin to show. so the idea of a practical clinical trial ispositing a question and conducting the trial as if you had the informationcollected specifically for the trial. it's a dicey use ofclinical records if you ask me,

but it's an interesting approach to makingbetter use of the clinical records than we have right now. so the take-away points fromthis conversation are two. one of them is that we have to helpthe initiative understand that clinical information is more than ehrs. and secondly, we have to help and expand theunderstanding that to construct knowledge of an individual may require bringing togetherknowledge sources from a number of different places, which takes us backto some of the earlier ideas, common ontologies,nomenclature that works, et cetera.

some brief comments on the frameworkfor community-based standards effort, which was the one that icompletely misunderstood as i went into it. i thought -- i read the word "community" tomean community like in out there as opposed to in here. it doesn't mean that at all. it means research community. i learned that quickly and i didn'thave to confess my lack of knowledge. but the basic thrust inthis conversation, again, was that the scientific communityshould determine the language standards,

the formatting standards, and the messagingstandards to bring this data together. so it was basically alogistical conversation. how do we define when this -- what we'recalling this phenomena and how -- where do we know the edge of the stomach iscompared to the beginning of the intestines, how do we define whatconstitutes an allele or a mutation? those kinds of things, theconversation said rightly so, the research committeeshould make those decisions. there was some tension between whati would call extractors and mappers. extractors are people who want the standardnomenclatures built first and have everyone

use them so they can thenbe extracted from them. the mappers say don't worryabout what people did let them do, let a thousand flowers bloom, let themdo whatever they want and we will map, we will link the term they useto the terms we want to use. this tension between extractors and mappersis both a philosophical and logistical tension, there are people who believe thatif we had enough computational power we could actually map everything wewouldn't have to have common terminology; that's not going to work. but there are people who believe commonterminology actually does represent a formal

and agreed upon labelingprocess, that also doesn't make sense. in the community standards, socialscience was a much bigger player, the university of michigan socialscience data repository was present, groups like phoenix;patients like me who are, i call perverse players in thefield, are coming in to say, we know things about peoplesexperience health every day. and we can bring this to the conversation sothat the community standard should not just be the biological and bioscience perspectivebut also the experiential human science perspective.

let me close with a coupleof words about ninr nursing, big data, and the patient experience. when we speak as nurses and wework with a phenomenon of nursing, working with a broad range ofindicators about an individual, largely my group, my training comes out ofthe social and physical sciences dimension, and not the biological science dimension soyou'll see me reflecting on this more from a social science model. but we deal largely withthe science and symptoms, the experience of individuals, what my grouphas been calling the observations of daily

living. these are the little pains that peopleget that alert them to health changes. they may not have any physiologic base butthe say something is different than you think it's going to be. images, though; we deal with lots of things,pictures of family dinners somebody walking down the hall; wombs [spelled phonetically]. we deal with data there of a non-reducibletype and often that makes the big data challenge more challenging. we deal with biomarkers certainly, andthose biomarkers are often proxies for some

underlying phenomenon, not exact indicatorsof that and that leads us to questions about accuracy and precision. we deal with things such as family dynamics,patient experiences over time that might be a femoral [spelled phonetically] mightdo away after a short period of time. the big data processes that i heard discussedpresume that the data exists and stays stable for a long period of time. much of what we deal with patientsdisappears after a day of conversation. population of phenotypes, we look at how acommunity is displaying a disease process and not an individual.

so we see a wide range of data that nevercame up in many of the conversations in the bd2k initiative. i'm coming here today to say these types ofdata have to come up in the conversation. and the reason why we need them have to bemotivated by the questions we are trying to answer. fundamentally, big data might advance nursingscience if we begin with a starting point that say the game is changing. there are new values, thereare new models of research, and there are new philosophies, andthe idea of precision and replication,

while important, is not enough to capitalizeon the new ways we have to understand the phenomenon of patients. if we think of big data as a strategy tounderstand the experience of individuals we will go a lot further in leveraging theinvestments that were coming into our needs. i have four recommendationsthat i'd like to put forward. id first like to plead -- and i have beensaying this for 26 years to this community, informatics training has tocomplement statistical training. we somehow have moved informatics into theelectronic health record or something that makes sure all ourseniors get before we leave.

informatics is the labeling,formalization, and management of data. that's all it is, and we do it for differentpurposes driven by where we are in our practice or our research careers and whatkind of phenomenon we are dealing with. but we often separateinformatics from statistics, believing that statistics somehow is dealingwith information that is known because we have actually labeled it and defined it andgiven an operational definition of it and therefore we know what it is. if we integrated informatics and statistics,we would make enormous steps just making all of our nurse researchers skillful inthe ability to use and make useful the

information that contributesto the big data initiative. team science is critical, and team science isnot just me and my students sitting together and having conversations unless my studentsare trained by six other disciplinarians also. we need to recognized that what we knowin nursing contributes to the health of a person; it is not the definition of thehealth of a person and by driving our research in a team science model we have thecapability of both appreciating what others do as well as making moreevident what it is that we do. embedded laboratorypartnerships are essential,

almost all of our key 232 programs andmany of our research training programs are at universities where they're already arei2b2 groups or large data centers and participating and collaborating with thosefrom a faculty level to the newest student is a strategy that we mustbe able to leverage more. there's lots of money,there's lots of investment, there's lots of talented people, and bringingthese together will help us know where big data can and may not be as useful to us. i want to bring a last callfor varied methodologies. we have a tendency and we have a traditionfor doing a good job with teaching our

students and preparing scientists inqualitative and quantitative methodologies that are reasonably well-known or canbe developed into cutting edge tools. the use of -- it's gone right out of my headnow -- survival analysis to determine when individual are going to experience problemsin a critical care unit is an example where we have done well with statistics,let's break the walls down and move out. we need to think about new kinds ofmethodologies like spaciotempo [spelled phonetically] dynamic modeling, some flexiblemodeling for example looking at complex patterns over time, it can take advantageof tools of engineering as well as business. mechanistic modeling, agent based modeling,these are all expertise that is held outside

of our discipline that could beleveraged for our discipline, and questioning. not that everyone needs to learn all these,but helping our students understand plurality of methodologies iscritical, and our journals, and our leadership. last, with attention to the work force ileave you with thinking about three critical points of development. we do need to have somenurse data scientists. we absolutely need some people who knowexactly where they are on that metro map and

know exactly how to get to another place. we more importantly need nurse scientistswith data sophistication and patients absolutely need dataintensive nurses in practice. we need to know what to do with that data tohelp the people we are trying to take care of. so i think there should befour types of investments, or four types of talentdevelopment investment, information at the point of care, creatingvisualizations that help a staff nurse overworked at a point of acomplicated patient know what to do.

and i saw on one of the iq questions: whatare the good strategies for a brand new nurse dealing with a patient who's crashing? this is a good idea, we need to thinkabout how to develop visualization tools. but also, how do we developefficient tools for data capture? because right now, what we do is we oftenhave every data point go through the nurse's eyes and fingertips and that won't scale. so we need to think aboutmonitors, body sensors, wearable computing, strategies that allowus to understand without actually touching. we need to consider information atthe point of knowledge generation.

the informatics that we need for datamanagement and data analytics have to stretch beyond what we know now. if an ro1 can keep a datamodel in the pi's head, it's too small. we just have to say we've got to reallyrecognize investments have to be leveraging larger questions, importantquestions with larger answers. workforce development; i mentionedcritical to have a plurality of thinking. remember that we as nurses are both consumersas well as contributors to big data and we have a reason to have a place at the tablebecause we know things that no one else does.

thank you very much for your time. [applause] >> patricia grady: thank you so much for thatenergizing and very informative presentation. the floor is open for questions. >> elaine larson: so patty, this whole thingis really scary because none of us are ready for this. and i thanks to lynn cardy i was able toparticipate in that first training workshop, and i have to say that there was such ahuge divide between the few people who were anything resembling clinicians or people whounderstood clinical sciences from the data

science people who thought that aside fromthe thought that just recommending medical students need to learn this stuff, the bigdata people seem to think that all they had to do was tell the clinical people what wasavailable and then we'd be able to work with them and figure out how to do it. but it's much more -- it's going to be hugelyexpensive and complicated so these are good suggestions, but whatare we all going to do to, you know, get our students and our juniorfaculty not to mention ourselves up to speed so we can even be at the table? >> patty brennan: i amthinking of three things,

one of them is inappropriate, andit was retirement sounds wonderful. [laughter] i worked with a guy about 20 yearsago from berkley whose name is escaping m, who said we are going to have ingestiblebiochips that we are just going to eat the knowledge that we haveand then we will know it, now that's not a bad idea. remember we know already, we knowalready, so they things that we know: ethics, inquiry, analysis, question answering, weknow and that gives us rights to be there. when we first begin to partner with acolleague and cardiovascular phenomenon or in public health, their languages are different,we already have the skills to say now what is

that mean to you, give that back to me. so there is aconversational set of skills we have. we need a cloak and an imprimatur, bygod you guys have to listen to us also, so it's a question ofleveling the conversation. and some of that, frankly, is going tocome by having enough people who have the technical terms down right whounderstand who can be the bridge makers, bridge walkers, who can,bridge builders maybe, who can engage with those fields. so some of it is embedding our students andour trainees in those groups even if we don't

know exactly why to start out. get over there, meet thosepeople, think about what they're doing, and tell us about it. more broadly as a field, our journals needto understand that these explorations in new areas has got to be a reason topublish, not a result of publishing. so we need to get our journals to be thinkingabout the main scripts of conversation, to open our ideas, to say here'sthe framing, here's the challenge, this is the approach are there other ideas. too often our journals want ouranswers, what are our answers,

we love that and jamie and i know that, so iam [unintelligible] of an informatics journal and we do the same thing. i don't know if all of us old people in theroom are ever going to get up to speed on all of this. i've stopped learning newprograms when n note came out; i just couldn't learnthat one, so that's gone. so one of the things we do needto do is embed ourselves in teams, and so right now i am working with a teamwhose work has been mostly funded by gm to build the onstar in your car and we areapplying it to asthma detection management.

the same thing that tells you that yourbattery is dying is now going to tell you that you are probably going to needyour rescue inhaler in two hours, we think -- we hope. so working with it on a project byproject basis which is an extremely long term strategy is something that i think iscritical when we look at our profiles for t32 trainers we need to lookat how diverse they are, not how coherent they are. we absolutely need leadership, we need topleadership to be constantly asking where is nursing giving andgetting to this all the time.

and its necessary public policyactivism around this is necessary, the investment they've put in the data setsthat are going to reflect one perspective on a patient and not another, can be modifiedif we have enough nurses saying we need these other elements for the electronic record. so i apologize for the rambling answerbut we've got to work on a couple different strategies at once. yes? >> female speaker: i was around when thebeginning of genetics came into nursing science and a lot of what i hear you talkingabout strategies that you are proposing are

very similar to what i saw as barriers andfacilitators incorporated to genetics and nursing science in 14, 15 years. do you have any thoughtson that -- for example, you were saying building teams, so a lot ofus might do genetics on our team but over time we had an inr who actually did thetwo months genetic courses and then it's now moved out to a lot of universitiesand genetic courses at universities. and you also talked about the younger peopleare going to be more intuitive about this and that's the same languagewe used about genetics, do you see similarities or bigdifferences as you think about them?

>> patty brennan: i really thank you for thisquestion because i appreciate the opportunity to reflect on it. i see similarities and i see differences. we had a compellingreason to understand genetics. we had a really compelling -- imean, there is just no question. this was a way to understand the levelof a person that we didn't have before. we don't have a compellingreason to understand big data; big data is an opportunity waiting to beexperienced rather than a resource solving a need.

that doesn't mean that we should dismiss it,but it does mean we need to think about it differently. so one of the things i will plea foris versatility as nurse investigators. i -- when carol was introduced and wasdescribed as having three ro1s at the same time, we think this is oddin nursing, cool go for i, girl, but really we treat it as unusualand we have to stop treating it that way. we have to have nursing involved in alot of research all the time and our team, we move in and out of teams, soour idea that we will create a team, for some groups it will still scale, idon't mean to be dismissing entirely.

but we need to be also think going about amodel where i am on this team for two years and then i move over to that teamand then i move over to that team. and that really sort of questions our conceptof long-term relationships that we really do build a lot more of in nursing. i think that in some ways genetics andgenomics place something in front of us that really complicated tothink about, really hard, and big data is and isnot complicated and hard. so the difference there to me iscomputational tools are difficult to learn, and not everyone will learn the tools,but there are resources that can help the

computation. what we have to help peoplelearn, which i think is much harder, is which to use and why. and like elaine said, the attitude right nowin the bb2 stations that i've been at on the campus has largely been,oh, we got all this data, call us up and tell us whatyou need, we can send it to you. and the idea that we needsomething at a level of granularity, over a period of time that maybe wasn'tthought of in a data set is not known so we have to find new ways tohave these conversations.

i think that the markers, or the payoffof having what we couldn't get of genetic knowledge is different and moreclear than the payoff of big data, and so we might need to think about anefficient strategy for our discipline as a whole, and that's where this body actuallyhas a critically important role in how do we balance the individual investigator-initiatedquestion of the moment, "i'm just curious," with building thecapacity to have a strong work force that is able to take on whatever will be in20 years what big data is for us now. you know, maybe it will be big smells, idon't know but there will always be something new.

so you track genomics. i was talking to bill about this earlier,this feels like 1987 with hiv research, nobody quite knew what to do, they knew theyhad to do something and we got -- we made progress both with boldsmart nurses as partnerships. so i don't think the strategiesare going to be all that different. the money is different now, the risks are alittle bit different and frankly i'm not sure if our resources are up to it yet. i don't think we've demandedour academic nursing programs, particularly phd programs, to have the levelof analytical skills and have our students

coming out with a levelof analytical thinking, particularly quantitative thinking, that willallow them to participate and that i am the most concerned about. and i'm not dissing qualitative researchbut we need to have our students taking math-based statistics,not applied statistics, we need to up the mathematicalsophistication of nursing overall. yes. >> kenton kaufman: thanksfor a great presentation. >> patty brennan: thank you.

>> kenton kaufman: a few practicalquestions and maybe one big question. if you are going to startaccessing the electronic medical records, how do you deal with hipaa? and how do you ensure qualitydata instead of garbage in, garbage out? and then the big question is are you talkingat all about accessing geospatial data say from satellites so you canlook at not just the u.s. but the world based population so you can tryand learn what's happening around the world and how it affects people?

>> patty brennan: those twoquestions are actually quite related, thank you very much for them. the first one had to do with how do you dealwith hipaa and the second one said what about geospatial data and various kinds of data? what we once called health data isno longer enough to describe health. so one of the ways we deal with hipaa is wedeal with the institution based risk-prone data within that institution. but what's on my cellphone is not hipaa protected. that's good and bad.

so we have to think about then how dowe ensure the quality of that data? some of the work that was done this summerwith the fda safety and innovations act risk base guidelines, risk baseregulatory guidelines are beginning to say, look at what the fda are doing to showthe integrity of information as it's used. a lot of what's happening in data exchangeright now in the health information exchange model, not the insurance exchanges but theinformation exchanges focus on providence and the way to ensure data quality. the issue of garbage in, garbage outto me is only one piece of the data, there are perverse motivations to emphasizeone type of data over another in health care

it's easier to focus onone kind than another. so we actually end up circumscribing ourdefinition of health problems to a level that might not be appropriate. and so i think we need to be looking ata more fundamental information's science framing for research activities wherewe begin to tolerate the fact that the information has anuncertainty imbedded in it. and most of our models don't allow forthat or we cut them off as outliers. so by that i mean when we areusing extracts from large data sets, we have to help people understand that andpeople need to have the skills to determine

the accuracy and the validity and the termthat the garter [spelled phonetically] uses the veracity that is this data point true fora long period of time or a short period of time. my temperature was elevatedwhen i had the flu two years ago; it's not elevated now. so that's not a good description of me now. so we are looking at time variant data. now to go to issues that aregeospatial or time bearing data, absolutely, the conversation touches onthe fact that health and health data,

health may be informed -- or health knowledgemay be informed by all sorts of things. so in fact your cell phone pings may actuallybe a really important aspect to indicate something about your personal health aswell as the health of your community. but the analytics and thequestions for that still aren't clear. so we see a lot of reallyinteresting exploratory things happening, looking at twitter feeds todetermine health concerns, the national library of medicine now isbeginning to mind some of the twitter feeds about their work to look at some of thehealth problems people want information about.

but the questions need to beasked before the data get looked at. so were at the situation right now, were likethe old men with the water water everywhere, nary a drop to drink -- youknow, we have a lot of data, the challenge is the same challenge we'vehad for 50 years since we've had moderned the scholarship in nursing. what questions are relevant and howwill we know we have answered them? oh gosh yes. >> female speaker:[unintelligible] [no audio] >> patty brennan: i want to stress alsothat the methods of delivering training,

and the methods of logicaltraining really are partners. so might learn of a methodthrough a training program, you may learn how to implementa method in a training program, but my investment, frankly, would be let'sstrengthen the nursing and partner for the methods as opposed to let's bringtoo much of the methods into nursing. i also want to ask those of you who areactive in the doctoral conference the aacn doctoral meeting to bring this conversationinto that conversation because very often the way we can affect curriculum isthrough the aacn and i don't see them, this is not on their radarscreen at all, as far as i can tell.

we need to remember thatinformatics is not just computer systems, and big data is not justlots and lots of stuff, it's questions about the humanresponse that were able to answer. and so that's the conversation that should begoing more to our doctoral training sessions than in this room. yes, bill. >> william holzemer: this concept of youmust have the priory question to me is -- i understand where you are coming from but italso seems to go against the early literature on data mining and discovering relationshipsthat would guide you to further testing

different kinds of things. so you seem very absolute on that... >> patty brennan: of course. [laughter] >> william holzemer: -- position -- so i wantto know where the dents in the armor are or so to speak. >> patty brennan: that's fair, i amdesperately terrified that our nursing work force shortage is worseat the investigator level. that is we are really shorton having great investigators.

so if i were going toask people to do anything, the data mining approach, they explore anddiscover and see what comes through i think is inefficient. so it's not a bad strategy if you knowwhy you want to do it in the first place, but if you just say i'vegot a data set i think will, so even the most not directed data mindersbegin with some general idea of what they are trying to answer. if they are trying to study cardiovascular ororthopedic problems or family dynamics versus self-concept.

so we should have a framing, becauseotherwise we are going to end up with a pile of data that just has to be sorted againand sorted again until we actually get to answering questions. i would like to incentify schools of nursingto really look at the portfolio of projects they are addressing, because not everybodyin a school or in a research think-tank or intermural institute needsto be doing the same thing, but each research intensive environment musthave some group that's doing something in the area of looking at largedata set, large data analysis, data integration from different sources,something so we can begin to fill in the

mosaic of big data in nursing. and i think that will only happen if weremain clear about what we want to know and that may be thedifference or you may call it a dent, i call it a feature. the question doesn't have to be so narrow,as to say if i don't get this answered i've wasted my time. the question needs to be more a statementof purpose a reason to continue to go on. a reason we should pay you to do this andnot somebody else who's solving the child who needs the conversation about the end of life.

i mean, gosh, that's a serious problemand were competing with them for resources. but we need to make sure that either wepartner around the significant problems in our research-intensive enterprises or wepartner with significant researchers across the country or across the street at otherschools to make sure we do not have to do everything ourselves. yes, i'm sorry do we need to stop? i'm having fun. >> female speaker: my child [unintelligible]. i teach statistics and i've been workingwith nurses for about six or seven years

[unintelligible] and i'm teaching in astatistics for nurse's course and i have tried to create the course sothey're data aspect is shown. so i would like to be involved just totalk with somebody about curriculum of this, i just think it's an interesting area. >> patty brennan: i think it's reallyimportant and i think that often we think of statistics as several method models andthat's it but in fact it's all where did your data come from what do you want to do with itwhat did you do with the errors and what did you do with the stuff you stuffed in thedrawers you know those are questions we have to answer so i'm glad to hear data managementwhich to me is where phlegmatic [spelled

phonetically] flows a lotcomes into the conversation. in 1995, connie and i made a proposal tosigma theta tau that we start sort of a data repository and sigma theta tau inindianapolis to common definitions or at least common definitions of terms nurses usein research because there were a lot of them, and it failed miserably. in part because it was underdeveloped andprobably not well written but it was also too early for people to think, i don't haveto create the universe myself in each data element that i define, i can use not justterms that have appeared in the literature but some terms that may appearin the electronic health record.

i cannot begin to think about the way i labelsomething can make not only my work better but make my data useful to someone else. >> patricia grady: thishas been really terrific, you have really stimulated us to think in newways and given us a lot of suggestions that we are going to follow andprobably get back to you on. thank you very much. [applause]