Fulltext.pdf

How do clinicians reconcile conditions and medications?The cognitive context of medication reconciliation Geva Vashitz • Mark E. Nunnally • Yisrael Parmet •Yuval Bitan • Michael F. O’Connor •Richard I. Cook Received: 17 April 2011 / Accepted: 22 August 2011Ó Springer-Verlag London Limited 2011 Medication omissions and dosing failures are subjects matched conditions and medications related to frequent during transitions in patient care. Medication the same organ system together (Wilcoxon W = 1917.0, reconciliation (MR) requires bridging discrepancies in a p \ 0.001). We conclude that the clinicians commonly patient’s medical history as a setting for care changes. MR arranged the information into two groups (conditions and has been identified as vulnerable to failure, and a clini- medications) and assigned an internal order within these cian’s cognition during MR remains poorly described in groups, according to organ systems. They also matched the literature. We sought to explore cognition in MR tasks.
between conditions and medications according to similar Specifically, we sought to explore how clinicians make criteria. These findings were also supported by verbal sense of conditions and medications. We observed 24 protocol analysis. The findings strengthen the argument anesthesia providers performing a card-sorting task to sort that organ-based information is pivotal to a clinician’s conditions and medications for a fictional patient. We cognition during MR. Understanding the strategies and analyzed the spatial properties of the data using statistical heuristics, clinicians employ through the MR process may methods. Most of the participants (58%) arranged the help to develop practices to promote patient safety.
medications along a straight line (p \ 0.001). They sortedmedications by organ systems (Friedman’s v2(54) = 325.7, Medical cognition Á Medical expertise Á p \ 0.001). These arrangements described the clinical Diagnostic reasoning Á Patient safety Á Card-sorting Á correspondence between each two medications (Wilcoxon W = 192.0, p \ 0.001). A cluster analysis showed that the A common fragile point in health care is transition betweenprovider and locale (Cook et al. 2000). A transition creates arisk for loss of information, abandonment of care plan, or discontinuity of treatment. Unintentional failures in medi- article (doi:10.1007/s10111-011-0189-0) contains supplementary cation prescribing are one of the most frequent causes of material, which is available to authorized users.
preventable harm in health care (Cornish et al. 2005; Bud- nitz et al. 2006). Discrepancies may arise from incomplete, Department of Industrial Engineering and Management, opaque, or ambiguous findings, complex medication inter- Ben-Gurion University of the Negev, P.O. Box 653, actions, and time pressure. Patients frequently do not con- 84105 Beer-Sheva, Israele-mail: gevava@bgu.ac.il vey valid clinical data in such transitions because ofmedical illiteracy, memory limitations, embarrassment M. E. Nunnally Á Y. Bitan Á M. F. O’Connor Á R. I. Cook (e.g., lifestyle medications and psychiatric medications), or the perception that information is clinically unimportant Department of Anesthesia and Critical Care,University of Chicago Hospitals, Chicago, IL 60637, USA Medication reconciliation (MR) can be broadly defined as the task of bridging discrepancies in a patient’s medicalhistory after a care setting changes. Linkages between conditions and medications should be explained. However,MR is commonly identified as vulnerable to failure (Clay In a simulation experiment, participants were asked to et al. 2008; Jylha and Saranto 2008; Miller et al. 2008; make sense of hypothetical conditions and medications Pippins et al. 2008; Wong et al. 2008; Brady et al. 2009; using a card-sorting task. Details of the experiment are Frei et al. 2009; Gandara et al. 2009). Awareness to MR described elsewhere (Vashitz et al. 2010). The card-sorting interlaces within the growing awareness to patient safety method is a validated method in the cognitive and social sciences for gathering user input (Coxon 1999). Analyses Landrigan et al. 2010). The Joint Commission, the of card-sorting tasks yield affinity diagrams that spatially accreditation body for US hospitals and other health care represent a subject’s concepts, mental models, or percep- organizations, has identified MR as a national patient safety goal (JCAHO 2006). This focus led to diverse efforts to Participants were clinicians in the Department of audit and improve the process of MR. Most of these efforts Anesthesia and Critical Care at the University of Chicago are prescriptive, such as the use of forms and emphasis on Medical Center, who practice MR daily. We abstracted supervision by pharmacists (Pronovost et al. 2003; Bo- patient records to produce a fictional case for preoperative ockvar et al. 2006; Hayes et al. 2007; Manning et al. 2007; assessment by an anesthesia provider. We chose the case Weingart et al. 2007; Coffey et al. 2009; Walker et al.
to replicate typical clinical complexity. We printed the 2009), or the use of information technology (IT) as a tool fictional patient’s diagnoses and medications on simple for data management (Poon et al. 2006; Kramer et al. 2007; paper cards. All participants faced the same initial card Turchin et al. 2008; Agrawal 2009; Schnipper et al. 2009).
arrangement shown in Fig. 1. The patient was described as However, the question of how MR relates to a clinician’s a 66-year-old woman scheduled for a wide local excision cognition of a patient’s medical history is currently unex- of a tongue lesion. We asked the participants to arrange the cards in a sensible way while ‘‘thinking aloud’’ and sharing An extensive body of literature explored cognitive pro- their thoughts. Three cards (cerebrovascular accident, cesses related to clinical diagnostic reasoning (Boshuizen clopidogrel, and digoxin) were exposed later in the simu- and Schmidt 1992; Patel et al. 1997, 2002; Charlin et al.
lation to assess the response to new data. For the methods 2000; Round 2001; Elstein et al. 2002; Thomas et al. 2008; described below, we analyzed the final arrangement of all Vickrey et al. 2010). MR can be described as a cogni- cards. A video camera captured hand movements and tive problem-solving task, similar to popular memory or matching games. MR requires memory, reasoning, andprioritization based on incomplete, ambivalent, or redun- dant data. Such functions place an extensive cognitive loadon the clinician, especially because MR is commonly per- We used geometric and spatial cues such as card order, formed under intense working conditions. MR may include alignment, and clustering to explore the perceptions of the typical characteristics of difficult problems, including time participants. We captured a graphic image of the final constraints, interactions between parts, uncertainty, and risk arrangements for each participant and coded the position of each card by its rectangular coordinates (x, y) in units of We previously demonstrated that clinicians performing pixels. Using these coordinates, we calculated the Euclid- a simulated MR task arranged medical conditions along a ean distance between each card pair. To make distances line ordered by organ systems (Vashitz et al. 2010). We comparable across participants, we standardized the raw were curious to further explore whether such patterns distances based on the longest distance in any particular appear with medications as well and whether clinicians reconcile conditions and medications in particular patterns.
Our specific aims were to (1) explore how the ordering patterns previously observed in medical conditions arereflected in the arrangement of medications and (2) to An important spatial measure of card sorting is alignment, describe the relationship between conditions and medica- as it may represent some communal property or even pri- tions and the way the relationship might help define MR in ority between cards. A linear order exists if the variance of practice. Such exploration is important to learn how cli- the coordinates projected on one of the axes (either X or Y) nicians make sense of medication and condition history and is smaller than the variance on the other axis. We used the how this may potentially improve patient safety.
Levene’s test for equality of variance to compare the Fig. 1 The simulated case.
CAD coronary artery condition,DVT deep vein thrombosis, variances on X- and Y-axis projections. This test was 2.2.3 Clustering conditions and medications applied to each participant separately. We then used anonparametric binomial test to test the hypothesis of a line- In our hypothesis, a shorter distance between a given like arrangement across all participants.
medication and a condition suggested a clinical relation.
We calculated the mean of the adjusted distances (MAD) of each condition–medication pair across all participants.
We then ran a cluster analysis on the MADs using 4 We were also interested in whether the cards were ordered clusters to replicate an ordinal scale: cluster 1, very close; in a meaningful pattern. One way to identify patterns in cluster 2, close; cluster 3, far; and cluster 4, very far. Pairs arrangement is to look at the proximity between cards. We in a same cluster might share clinical properties. We used the Friedman test for ranking to compare the adjusted compared the MAD to the classification by the senior cli- distances across participants. The Friedman test yields a nicians using a nonparametric two independent samples mean rank for each card pair, which is the average of the pairs’ ranks across all participants. The smaller mean rankfor a card pair indicates that, across all participants, these cards were closer. We compared the mean ranks in thecontext of the clinical classification of the conditions and We aimed to explore ‘how’ clinicians reconcile the infor- medications according to the organ systems treated. We mation by analyzing the think-aloud verbal protocols and used the nonparametric k independent samples test to cor- the post-experiment interview. We sought to identify relate the Friedman’s mean ranks with the classifications of qualitative terms that may explain underlying cognitive senior clinicians. The classification was performed inde- process. The analysis focused on explanations about the pendently by two senior clinicians (MN, MO), who were way the cards were sorted, such as sorting criteria, order of unaware of the mean ranks. If two medications affected the same organ system (e.g., pulmonary and pulmonary, psy-chiatric and psychiatric), we referred to them as a ‘‘match’’.
Such matches reflect the influence of clinical reasoning on card sorting. We compared the mean rank of each medi-cation–medication pair, according to whether it was a The participants sample has been described in detail ‘‘match’’ or not by the classification of the senior clinicians previously (Vashitz et al. 2010). We recorded results using a nonparametric two independent sample test.
from 24 participants: 6 attending physicians, 5 certified registered nurse anesthetists, 10 residents, and 3 third-year p \ 0.001). All the pairs in cluster 1 (very close pairs) were medical students. Ten participants were women and 14 matches, as were 59.3% in cluster 2 (close) and 39.5% in cluster 3 (far). There were no matches in cluster 4 (veryfar). Participants tended to match conditions and medica- 3.1 Alignment and proximity between cards tions related to similar organ systems together. For exam-ple, cardiovascular conditions (atrial fibrillation, coronary The Levene’s test for equality of variance showed that 14 artery condition, hypertension, myocardial infarction, and participants (58%) arranged the medications along a cerebrovascular accident) are treated by cardiovascular straight line (p \ 0.001). The 11 medications yielded 55 medications (aspirin, atorvastatin, clopidogrel, digoxin, medication–medication pairs for each participant. Table 1 diltiazem, and potassium). With some exceptions, the demonstrates the closest and farthest pairs and whether analysis put them into the same cluster, as it did with the they are a clinical match. All pairs are available from the online appendix. Friedman’s mean ranks were significantlydifferent from each other (Friedman’s v2(54) = 325.7, p \ 0.001). Lower mean ranks describe a pair of medica-tions that were placed closely and presumably belong to a Analyzing the verbal protocols, we looked at both the same organ system. The nonparametric two independent ‘‘think-aloud’’ portion of the task and the post-experiment sample test showed that the mean rank in ‘‘matched’’ pairs reflections. Many subjects mentioned sorting the cards by (10.1) was significantly lower than the mean rank in ‘‘unmat- ched’’ pairs (37.4) (Wilcoxon W = 192.0, p \ 0.001). In otherwords, the participants tended to sort medications treating …It’s helpful for me to kind of think about it from either kind of an organ system approach (Subject 14).
…I think my first inclination is to kind of group these 3.2 Clustering conditions and medications relations in terms of anatomical location or patho-physiology(Subject 4).
The cluster analysis classified the relationships into four …I guess its kind of how we learned in medical groups based on proximity, around centroids at MAD of school, first you have your history or your present 0.40, 0.48, 0.55, and 0.64 (Fig. 2). We expected that clin- ically associated conditions and medications would be The subjects also mentioned pairing medications with placed in a similar cluster (i.e., condition–medication pairs that were related to a same organ system). We ratified theserelationships statistically by comparing the MAD to the …I organize things according to the disease states…I classification by senior clinicians using a nonparametric feel it is incumbent upon the practitioner to make sure two independent sample test. Consistent with our hypoth- that a medication correlates with at least one diag- esis, the MADs correlated with the clinical match between nosis that we know the patient to have. I tend to lump conditions and medications (Wilcoxon W = 1,917.0, things into systems, organ systems… (Subject 15).
Cardiovascular or neurological/cardiovascular Cardiovascular or neurological/psychiatric is the average of the pairs’adjusted distances across all Cardiovascular or neurological/psychiatric Psychiatric/cardiovascular or neurological in-group orders. If both conditions and medications arearranged by organ-based order, the linkages between conditions and medications should have a similar rea-soning pattern. Data from the verbal protocols and thepost-experiment interview support the findings from thequantitative analysis. These data suggest that many sub- jects sorted the cards by organ systems and matchedmedications with conditions.
4.1 Cognitive insights reinforced by findings Each card had various attributes, such as priority, relevance to the forthcoming procedure, possible links with other Mean Adjusted Distance (MAD)
cards, and time of occurrence. Such attributes may dictatethe sorting strategy. Different clinicians may distinguishand weigh such attributes differently. The clinicians apparently used the attributes to classify cards by com- munal properties: it is clear that the clinicians sorted the cards into two groups (conditions and medications) andassigned an internal order within these groups according to Fig. 2 Cluster analysis of relationship between conditions and organ systems. They also matched cards according to medications. A cluster is a group of condition–medication pairs withan adjacent mean of adjusted distances (MADs). The X-axis depicts a related criteria (conditions that are usually treated by cer- nominal number assigned to the pair (from 1 to 110). The Y-axis tain medications, and medications that usually treat certain depicts the MADs of each pair across all participants. Shorter MADs conditions). Such strategies are not obvious because other represent cards that are closer together. A ‘‘match’’ is a case in which clinical, chronologic, causal, or contextual criteria could the condition and medication belong to the same organ system (e.g., apulmonary condition and a pulmonary medication). The horizontal have been used. These relationships are based on a con- lines represent the center of clusters (centroids). Legend: Cluster 1: ceptual understanding of condition physiology. Such a filled circle matches, open circle non-matches. Cluster 2: filled square consistent trend probably reflects disciplines learned in matches, open square non-matches. Cluster 3: filled triangle matches, medical training, in which preclinical courses are often open triangle non-matches. Cluster 4: filled diamond matches, opendiamond non-matches …Then basically the main organizational scheme 4.2 Correspondence with clinical reasoning was just pairing the drugs with what condition they were likely to, to be… (Subject 24).
…I’ve got it laid out so the diseases are over here and 4.2.1 The ‘‘small worlds’’ concept the medicines associated with them are over here, sothere’s kind of a correspondence between them… Clinicians making diagnoses use reasoning through a network of causal rules that appears to derive from thephysicians’ underlying knowledge base, adapted to goalsof clinical tasks (Patel and Groen 1986; Charlin et al.
2000). Our findings suggest that medications and condi-tions share complex cognitive relationships, which clini- Our initial quest was to describe how clinicians make cians use during clinical work. The cognitive literature sense of the MR task. Our previous findings (Vashitz offers several explanations for such mechanisms. For et al. 2010) suggest a highly repetitive pattern of example, Kushniruk et al. (1998) showed that clinicians arranging medical conditions in an organ-based order.
organize diagnostic knowledge by similarities between The current data replicate this finding with medications condition categories, forming ‘small worlds’ consisting of and moreover suggest linkages between the conditions small subsets of conditions and their distinguishing fea- and medications. The integration of our previous and current data depicts a pattern of separating conditions and grouped together conditions and medications sharing medications into two groups, ordering elements of each communal features according to the presence of key group by organ system, and creating linkages based on 4.2.2 Family resemblance and representativeness Rosch and Mervis (1975) suggested that categories are not The simulated case was based on a real-life, complex organized around strict definitions but rather according to a preoperative evaluation that reflects a task clinicians face family resemblance. Objects belong to the same category routinely. The experiment was conducted with minimal because they are similar to each other and dissimilar to instructions to allow spontaneous behavior by clinicians objects in contrasting categories. Ahn and Medin (1992) who practice MR daily. The sample included a range of suggested that people first arrange data by a preferred expertise, including senior attending clinicians, residents, criterion. If examples do not fit within the preferred crite- advanced practice nurses, and medical students. We rion, people then adjust for differences between preferred translated observed behaviors into quantitative data to and other criteria. Such an adjustment may represent a compromise between a structured concept and the neces- We acknowledge several limitations of our findings. Our sity of mapping concepts into real-world examples. Vari- simulation was for one patient and included clinicians of ability in our data probably reflects such an adjustment.
the same specialty. Sample size limited the exploration of Some cards in our study fit several categories. For example, variability and different expertise levels. Although we aspirin may be used to treat both cardiovascular and neu- think that the simulated case has a high fidelity to a real rological conditions. Deep vein thrombosis (DVT) can be patient, performance in an experiment might be different categorized as a cardiovascular or a hematologic condition.
from care of a real patient. Our observations do not address Hypertension is a risk factor for cardiovascular conditions, ambiguities and conflicts clinicians might encounter when which can lead to myocardial infarction. Depression in the performing MR. Exploring the findings with various clin- simulated case may have resulted from cancer or a heart ical cases, specialties, and expertise levels may uncover condition. Less-connected cards potentially reflect uncer- ambiguities and conflicts at the heart of MR and advance Several studies (D’Zurilla and Goldfried 1971; Rath et al. 2004) suggested that problem-solving usually begins with a general orientation or ‘‘set’’, followed by variouscognitive-behavioral steps, which ideally lead to effective The insights into the thought processes of clinicians problem resolution. The distinction between conditions and during MR are a starting point for discussions about what medications appears to underlie a general orientation, fol- makes medical care safe or vulnerable. The strategies lowed by arrangement by organ systems.
identified here may serve to understand underlying cog- The categorization also may be derived from a represen- nitive processes. The results of the study can be reused tativeness heuristic (Tversky and Kahneman 1974). Cards for the purpose of providing clinicians with decision aids may be classified into a category because they saliently that support these patterns. Efforts to improve safety represent it. For example, a deep venous thrombosis (DVT) should strive to replicate the natural thought process of is a hematologic representative, but it may be categorized clinicians. Our findings support the argument that for MR into other groups as well. Whether such strategies were used, safety, organ-based information should be considered and their temporal order, should be further explored.
pivotal to a clinician’s cognition. We suggest that suchtools should be aware of these strategies and assist cli- nicians with forming clinical linkages between conditions and medications. Such tool may follow previous conceptsof graphic user interface with anatomical diagrams used MR is apparently an interplay between long-term concep- to facilitate medical information gathering and entering tual models of anatomy and medications and a working- (Stoicu-Tivadar and Stoicu-Tivadar 2006). MR should memory, problem-solving capability. Moreover, it is an apparently be approached as a piece of a larger organi- interplay between external representations (e.g., diseases zational, clinical, and cognitive process. Hence, it may be and medications) and internal representations (e.g., the integrated in broad interventions to improve safety, clinical reasoning that matches between diseases and including training, artifact design, and IT, for adaptive medications) (Richardson and Ball 2009). This insight would indicate what type of intervention could improve the The methodology may be applicable to other disciplines, effectiveness of the MR process and its reliability. As teams, technologies, and socio-technical contexts. The working memory is limited, we may suggest intervention methodology is independent of the discussed context and is that reduces working-memory workload, such as automated applicable to any other cognitive information gathering decision aids that map to observed cognitive processes.
2006 Society of Hospital Medicine national meeting. J HospMed 3:465–472 Coffey M, Cornish P, Koonthanam T, Etchells E, Matlow A (2009) We sought to explore whether ordering patterns previously Implementation of admission medication reconciliation at two observed in organizing conditions are replicated in the academic health sciences centres: challenges and success factors.
sorting of medications, and how clinical reasoning affects the cognitive relationship between conditions and medica- Cook RI, Render M, Woods DD (2000) Gaps in the continuity of care and progress on patient safety. BMJ 320:791–794 tions. The majority of the clinicians performing a medica- Cornish PL, Knowles SR, Marchesano R, Tam V, Shadowitz S, tion reconciliation task matched conditions and medications Juurlink DN, Etchells EE (2005) Unintended medication treating the same organ systems. The arrangements reflec- discrepancies at the time of hospital admission. Arch Intern ted a clinical reasoning between conditions and medica- Coxon APM (1999) Sorting data: collection and analysis. Sage, tions. These findings corroborate our previous findings and strongly suggest that medications and conditions share com- D’Zurilla TJ, Goldfried MR (1971) Problem solving and behavior plex group relationships that are likely used by clinicians modification. J Abnorm Psychol 78:107–126 to build cognitive strategies, using their own conceptual Elstein AS, Schwartz A, Schwarz A (2002) Clinical problem solving and diagnostic decision making: selective review of the cogni- understanding of condition physiology. The findings sup- port the argument that organ-based information is central to Frei P, Huber LC, Simon RW, Bonani M, Luscher TF (2009) a clinician’s cognition while performing MR. These com- Insufficient medication documentation at hospital admission of mon strategies are a starting point for defining MR. Such cardiac patients: a challenge for medication reconciliation.
J Cardiovasc Pharmacol 54:497–501 exploration is important to learn how clinicians make sense Gandara E, Moniz T, Ungar J, Lee J, Chan-Macrae M, O’Malley T, of medication and condition histories. An understanding of Schnipper JL (2009) Communication and information deficits in such perceptions may produce organizational strategies patients discharged to rehabilitation facilities: an evaluation of that fit and support the process and potentially improve five acute care hospitals. J Hosp Med 4:E28–E33 Hayes BD, Donovan JL, Smith BS, Hartman CA (2007) Pharmacist- conducted medication reconciliation in an emergency depart-ment. Am J Health Syst Pharm 64:1720–1723 This work was kindly supported in part by a Joint Commission on Accreditation of Healthcare Organizations Fulbright doctoral dissertation research scholarship to Geva Vashitz.
(JCAHO) (2006) Using medication reconciliation to prevent We thank Christine Jette, MD, and Annette Martini, MD, for their errors. Jt Comm J Qual Patient Saf 32:230–232 help in constructing the experiment. GV was supported by a Fulbright Jylha V, Saranto K (2008) Electronic documentation in medication doctoral dissertation research scholarship.
reconciliation—a challenge for health care professionals. ApplNurs Res 21:237–239 Kramer JS, Hopkins PJ, Rosendale JC, Garrelts JC, Hale LS, Nester TM, Cochran P, Eidem LA, Haneke RD (2007) Implementationof an electronic system for medication reconciliation. Am J Kushniruk AW, Patel VL, Marley AA (1998) Small worlds and Agrawal A (2009) Medication errors: prevention using information medical expertise: implications for medical cognition and technology systems. Br J Clin Pharmacol 67:681–686 knowledge engineering. Int J Med Inform 49:255–271 Ahn WK, Medin DL (1992) A two-stage model of category Landrigan CP, Parry GJ, Bones CB, Hackbarth AD, Goldmann DA, Sharek PJ (2010) Temporal trends in rates of patient harm Boockvar KS, LaCorte HC, Giambanco V, Fridman B, Siu A (2006) resulting from medical care. N Engl J Med 363:2124–2134 Medication reconciliation for reducing drug-discrepancy adverse Manning DM, O’Meara JG, Williams AR, Rahman A, Myhre D, events. Am J Geriatr Pharmacother 4:236–243 Tammel KJ, Carter LC (2007) 3D: a tool for medication Boshuizen HPA, Schmidt HG (1992) On the role of biomedical discharge education. Qual Saf Health Care 16:71–76 knowledge in clinical reasoning by experts, intermediates and Miller SL, Miller S, Balon J, Helling TS (2008) Medication reconciliation in a rural trauma population. Ann Emerg Med Brady AM, Malone AM, Fleming S (2009) A literature review of the individual and systems factors that contribute to medication Patel VL, Groen GJ (1986) Knowledge based solution strategies in errors in nursing practice. J Nurs Manag 17:679–697 Budnitz DS, Pollock DA, Weidenbach KN, Mendelsohn AB, Patel VL, Groen CJ, Patel YC (1997) Cognitive aspects of clinical Schroeder TJ, Annest JL (2006) National surveillance of performance during patient workup: the role of medical exper- emergency department visits for outpatient adverse drug events.
tise. Adv Health Sci Educ: Theory Pract 2:95–114 Patel VL, Kaufman DR, Arocha JF (2002) Emerging paradigms of Cacciabue PC, Vella G (2010) Human factors engineering in cognition in medical decision-making. J Biomed Inform healthcare systems: the problem of human error and accident Pippins JR, Gandhi TK, Hamann C, Ndumele CD, Labonville SA, Charlin B, Tardif J, Boshuizen HP (2000) Scripts and medical Diedrichsen EK, Carty MG, Karson AS, Bhan I, Coley CM et al diagnostic knowledge: theory and applications for clinical (2008) Classifying and predicting errors of inpatient medication reasoning instruction and research. Acad Med 75:182–190 reconciliation. J Gen Intern Med 23:1414–1422 Clay BJ, Halasyamani L, Stucky ER, Greenwald JL, Williams MV Poon EG, Blumenfeld B, Hamann C, Turchin A, Graydon-Baker E, (2008) Results of a medication reconciliation survey from the McCarthy PC, Poikonen J, Mar P, Schnipper JL, Hallisey RK et al (2006) Design and implementation of an application and Turchin A, Hamann C, Schnipper JL, Graydon-Baker E, Millar SG, associated services to support interdisciplinary medication recon- McCarthy PC, Coley CM, Gandhi TK, Broverman CA (2008) ciliation efforts at an integrated healthcare delivery network.
Evaluation of an inpatient computerized medication reconcilia- tion system. J Am Med Inform Assoc 15:449–452 Pronovost P, Weast B, Schwarz M, Wyskiel RM, Prow D, Milanovich Tversky A, Kahneman D (1974) Judgment under uncertainty: SN, Berenholtz S, Dorman T, Lipsett P (2003) Medication heuristics and biases. Science 185:1124–1131 reconciliation: a practical tool to reduce the risk of medication Vashitz G, Nunnally M, Bitan Y, Parmet Y, O’Connor M, Cook RI (2010) Making sense of diseases in medication reconciliation.
Rath JF, Langenbahn DM, Simon D, Sherr RL, Fletcher J, Diller L (2004) The construct of problem solving in higher level Vickrey BG, Samuels MA, Ropper AH (2010) How neurologists neuropsychological assessment and rehabilitation. Arch Clin think: a cognitive psychology perspective on missed diagnoses.
Richardson M, Ball L (2009) Internal representations, external Walker PC, Bernstein SJ, Jones JN, Piersma J, Kim HW, Regal RE, representations and ergonomics: towards a theoretical integra- Kuhn L, Flanders SA (2009) Impact of a pharmacist-facilitated tion. Theor Issues Ergon Sci 10:335–376 hospital discharge program: a quasi-experimental study. Arch Rosch E, Mervis CB (1975) Family resemblances: Studies in the internal structure of categories. Cogn Psychol 7:573–605 Weingart SN, Cleary A, Seger A, Eng TK, Saadeh M, Gross A, Round A (2001) Introduction to clinical reasoning. J Eval Clin Pract Shulman LN (2007) Medication reconciliation in ambulatory oncology. Jt Comm J Qual Patient Saf 33:750–757 Schnipper JL, Hamann C, Ndumele CD, Liang CL, Carty MG, Wong JD, Bajcar JM, Wong GG, Alibhai SM, Huh JH, Cesta A, Pond Karson AS, Bhan I, Coley CM, Poon E, Turchin A et al (2009) GR, Fernandes OA (2008) Medication reconciliation at hospital Effect of an electronic medication reconciliation application and process redesign on potential adverse drug events: a cluster- randomized trial. Arch Intern Med 169:771–780 Woods DD, Hollnagel E (1987) Mapping cognitive demands in Stoicu-Tivadar L, Stoicu-Tivadar V (2006) Human-computer inter- complex problem-solving worlds. Int J Man Mach Stud action reflected in the design of user interfaces for general practitioners. Int J Med Inform 75:335–342 Thomas RP, Dougherty MR, Sprenger AM, Harbison JI (2008) Diagnostic hypothesis generation and human judgment. PsycholRev 115:155–185

Source: http://www.ctlab.org/documents/How%20do%20clinicians%20reconcile%20conditions%20and%20medications.pdf