j ourna l ho me pa g e: www.elsev ier .com/ locate /hea l thpol
A framework to analyze hospital-wide patient flow logistics:Evidence from an Italian comparative study
Stefano Villaa,b,∗, Anna Prenestini c,d, Isabella Giusepic
a Department of Management, Catholic University, Rome, Italyb CERISMAS, Research Centre in Health Care Management, Catholic University, Milano, Italyc CERGAS, Center for Research on Health and Social Care Management, Bocconi University, Milano, Italyd SDA Bocconi, School of Management, Via Roentgen, 1, 20136 Milano, Italy
a r t i c l e i n f o
Article history:Received 7 February 2012Received in revised form23 December 2013Accepted 24 December 2013
Keywords:HospitalPatient flowsAnalysis framework
a b s t r a c t
Through a comparative study of six Italian hospitals, the paper develops and tests a frame-work to analyze hospital-wide patient flow performance.
The framework adopts a system-wide approach to patient flow management and is struc-tured around three different levels: (1) the hospital, (2) the pipelines (possible patientjourneys within the hospital) and (3) the production units (physical spaces, such as oper-ating rooms, where service delivery takes places).
The focus groups and the data analysis conducted within the study support that the modelis a useful tool to investigate hospital-wide implications of patient flows.
The paper provides also evidence about the causes of hospital patient flow problems.Particularly, while shortage of capacity does not seem to be a relevant driver, our data shows
Operations managementLogistics performance
that patient flow variability caused by inadequate allocation of capacity does represent akey problem. Results also show that the lack of coordination between different pipelinesand production units is critical. Finally, the problem of overlapping between elective andunscheduled cases can be solved by setting aside a certain level of capacity for unexpectedpeaks.
1. Introduction
In the last years, practitioners and academics have beenincreasingly paying attention to the control of hospitalpatient flow logistics. The purpose of patient flow logisticsis to optimize the management of patient flows through thevarious hospital production units (such as the emergency
department, operating rooms, hospital beds or outpatientclinics) from the patient’s arrival at the hospital to theirdischarge and first follow-up [1–3].
∗ Corresponding author at: Catholic University, Department of Man-agement, Largo F. Vito 1, 00168 Rome, Italy. Tel.: +39 339/1234.643.
Using empirical studies, several authors [1,4–11] havereported that failure to manage patient flows is the originof several typical hospital problems. These issues include:(i) queues and delays, (ii) under- and over-capacity uti-lization, (iii) variability of workload and stress for hospitalpersonnel, (iv) errors and (v) placement of patients in inap-propriate settings.
The current trend of developing hospital models thatare organized around processes and patients – to improvethe efficiency and quality of care provided – also makespatient flow management a relevant issue. In fact, if the goalis to overcome the traditional hospital model structuredaround clinical specialties, good management of patient
flow throughout the new multidisciplinary hospital pro-duction settings becomes critical [3,5,12–16].
Despite this growing interest toward patient flow man-agement, scientific studies have not yet developed a
A review of the scientific literature shows there are twodifferent types of studies that have approached the issue of
S. Villa et al. / Health
horough and system-wide framework to measure over-ll hospital patient flow performance. The creation of such
model is necessary in order to (i) compare the perfor-ance of different hospitals and (ii) identify the root causes
f hospital patient flow problems.In order to fill this gap, this paper (i) presents a model
o measure hospital patient flow performance and (ii) useshe model to test some explanations about the causes ofospital patient flow problems found in scientific litera-ure.
In order to meet this twofold goal, we adopted a compar-tive case method approach which is extensively explainedn Part 2. Part 3 presents a summary of main findings fromhe scientific literature to position our contribute and todentify the requirements for our model. The model is pre-ented in Part 4, while Part 5 provides the results of theest of the model to the hospitals included in the study.he last section is dedicated to outlining the implicationsf the results for managers, policy makers and researchers.
. Research goals and methods
As stated above, the main goal of the study is to developnd test an analytical framework to measure the perfor-ance of hospital patient flow logistics.The one-year long research design was characterized by
hree different steps as summarized in Box 1.The first step of the research protocol was an extensive
iterature review. Specifically, we carried out a literatureeview on MEDLINE, BSC, CILEA and EMERALD using differ-nt combinations of the following keywords: patient flow,ogistics, operations management, hospital and health care.
This first step was finalized to check the state of thert of the scientific debate on this issue and identify theost relevant patient flow dimensions to start building a
omprehensive and sound framework to measure hospital-ide patient flow logistics performance.
In this sense, the literature review helped the authorso frame a first theoretical model that was then necessaryo test and validate through a sound research protocol.
Given this guiding research question and the lack oftudies, a comparative case method approach was judgedhe most appropriate [17,18] in order to gain a deep insightnto patient-flow issues and develop a coherent and use-ul theoretical model. We eventually selected an overallumber of six cases, a sample size that turned out to beppropriate for reaching a certain level of theoretical satu-ation [19] in order to have enough material to develop theodel and test in real settings.Cases were chosen for theoretical, and not statistical,
easons [20]. To be included in the study, hospitals neededo meet three inclusion criteria [18]. First, it was necessaryo select cases where patient flow logistics was centraln the organizational strategy. The hospitals included inhe study are the first movers on these issues in Italy;hey inserted patient flow logistics as a priority in theirtrategic plans and, thereafter, they are involved in a
eries of relevant improvement projects. Second, hospitalsad to have already in place a system to collect patientow data and such data were (or were planned) to beublicly available. Finally, hospitals needed to be willing
15 (2014) 196–205 197
to identify four representatives, particularly: a physicianworking as part of the staff to the Chief Medical Officer,a nurse manager, a practicing physician and a controlleror operations manager. These four representatives wererequired to have direct responsibility over hospital patientflow management and willing to dedicate time to theproject by supporting the team in the data analysis andparticipating in different focus groups.
The second phase of the research (see Box 1) was charac-terized by a series of focus groups (six overall in one-yearperiod) with all the representatives from each of the sixhospitals. This second stage of the research protocol wasaimed at two major objectives. The first was to validatethe theoretical model. The dimensions that emerged fromthe literature review were discussed through an iterativeprocess conducted using the Delphi methodology and Lik-ert scale surveys from 0 to 5 to assess the relevance of thedifferent dimensions identified in the scientific literature[21]1.
The second objective was to define indicators and toolsto support the application of the model.
With the help of the focus groups, we identified the finalset of indicators to include in the model, selected the qual-itative tools to investigate the different hospital patientflows, and developed a standardized data collection pro-tocol.
It is worth noting that at the end of this iterative process,all the participants confirmed the validity and appropriate-ness of the model to perform a comprehensive appraisal ofhospital-wide patient flow logistics.
The third and last step of the project was aimed at: (i)testing the appropriateness of the model and (ii) verifying,with empirical data, the most relevant explanations (foundin the scientific literature) about patient flow problems.
In order to test the appropriateness of the model we firstmapped, at each hospital site, the most relevant hospitalpatient flows using the qualitative tools identified dur-ing the focus groups, particularly we used an activity- andactor-oriented flow chart and ISO (International StandardOrganization) symbols. Then, using a standardized datacollection protocol, we computed, for all the hospitalsincluded in the study, the various indicators emerged fromthe focus groups.
For all six hospitals, we analyzed approximately oneyear of data (from January 15, 2009, to December 15, 2009)comprising 335 observations. We omitted the observationsfrom the beginning and the end of the year to exclude thosepatients whose hospital stays had overlapped two separateyears.
1 In order to identify the most relevant dimensions to be included inthe framework we used the hospital representatives (four for each casestudy) as a practice expert panel. Particularly, we asked to evaluate, on aLikert scale from 0 to 5, the relevance and frequency of about 20 differentsources of patient flows problems and to discuss these results.
198 S. Villa et al. / Health Policy 115 (2014) 196–205
Box 1: Phases, methodology and goals of the study.Phase Methodology Goal
1. Development Literature review Identify the literature gaps and the most relevant patient flow dimensions to be included in thetheoretical modelDevelop a first proposal of the theoretical model
2. Validation Focus groups Assess the relevance of the different dimensions identified in the scientific literatureIdentify the specific indicators, qualitative tools, and collection data protocol to include in the model
ness of
nt expla
3. Test Interviews data analysis Test the appropriate
Test the most releva
patient flow logistics. First, there is a vast body of literaturethat, drawing from the operations research field, focuses onthe scheduling and optimization of single production units.Second, there is a more recent wave of operations man-agement studies that propose general theories and modelsabout the actual causes of hospital patient flow problems.
The operations research studies [22–32] focus partic-ularly on scheduling and planning issues in relation tosingle production units, such as the operating room or theemergency or radiology departments. These studies, usingsophisticated quantitative tools (such as queuing theory,simulation and linear programming), provide suggestionson how to maximize or minimize a given objective func-tion.
The current study is mainly built upon the evidence andresults of the second stream of contributions. In fact, weadopted the concept used by other authors [13,27–29,33]of the hospital as a complex system made of several inter-nal sub-components that are highly interdipendent witheach other. To create value in health care, it is critical toaddress the overall cycle of care, from the very first access tothe discharge and follow-up. Partial approaches that focusonly on single hospital resources or on discrete interven-tions, rather than the full production cycle, tend to lead theoverall system to suboptimal results. Consequently, mean-ingful and effective patient flow strategies need to adopt asystem-wide approach.
Several authors have proposed different theories andmodels to explain the actual source of hospital patient flowproblems.
One relevant aspect outlined by different authors[5,9,11] is that patient flow problems occur when hospitalresources (such as beds, operating rooms and nursing staff)are allocated in a fixed way and not updated on a regularbasis, thus reflecting the distribution of the historical rightsof the practitioners rather than the actual patient-relateddemands.
Vissers identifies three different factors that causepatient flow problems [34,35]: (i) capacity shortage: theamount of a resource available to a production unit maynot be in balance with the demand for that resource atthe average level of production; (ii) patient flow variabil-ity: the timing of the allocation of resources to a unit maylead to peaks and troughs; for example, if operating roomsessions are not well allocated during the week, large vari-ations in the demand for beds and nursing staff may result;
and (iii) lack of coordination: the capacities of the vari-ous resources that are required simultaneously by differentspecialties may not be balanced, resulting in bottlenecks orunder-utilization.
the modelnations (found in the scientific literature) about patient flow problems
The issue of patient flow variability has also been exten-sively analyzed by other authors [9,34–39]. Particularly,Litvak makes a distinction between natural variabilityand artificial variability [10,11,36]: natural variability isuncontrollable and is due to the intrinsic characteristicsof health care delivery (for example, patient flow fromthe ED), whereas artificial variability is potentially control-lable through managerial intervention and is due to processdefects or incorrect behaviors.
According to Litvak [11], the key to improving patientflow management – and, consequently, accommodategrowing demand – is to increase the bed occupancy byeliminating patient flow variability.
Finally some studies [40,41] stress the importance, inpatient flow strategies, of looking at “throughput time”,defined as the amount of time an item spends in the system,in order to identify the bottleneck, which is the limiting fac-tor stopping the patient flows from going smoothly throughthe healthcare pathway.
4. The analysis framework
The analysis of the scientific literature and the results ofthe focus groups identify six different possible main causesof patient flow problems, particularly: (i) bad allocation ofcapacity, (ii) shortage of capacity, (iii) variability, (iv) lackof coordination between the different hospital productionunits (e.g. operating room, ICU or regular floors), (v) pres-ence of bottlenecks along the whole hospital chain thatdelay the patient throughput and (vi) overlapping betweenelective and emergency/urgency cases.
In order to realize a comprehensive appraisal of hospitalpatient flow performance, we have tried to incorporate allthese dimensions in our theoretical model. Specifically, wedeveloped an analytical framework organized around threedifferent levels (cf. Fig. 1) where each level of the modelfocuses on one or more of these dimensions.
4.1. Hospital-wide system
The study compared the six hospitals on a series ofstandard indicators, such as bed-occupancy rate, numberof bed turns, percentage of emergency cases and averagelength of stay.
In this case, the main goals are to contextualize thepatient flow analysis within the general hospital contextand to determine whether there are situations of capacityshortage that create patient flow problems.
S. Villa et al. / Health Policy 115 (2014) 196–205 199
Hospital
Hospital Pipelines
Production Units
POSSIBLE INDICATORS FOCUS ON
• Beds occupancy rate • Length of stay • Bed turns • % of urgent cases • Distribution of cases between the different pipelines
- Contextualize the patient flows analysis within the general hospital setting - Understand if there are problems of shortage of capacity
• Census • N. of admissions • Length of stay • % of “direct line” patients
- Understand patient flow variability (across the day, the week and the year) through typical variability indicators (standard deviation, range, variance coefficient, ecc.) - Analyze the transfers of patients between pipelines - Understand the interactions between elective cases and urgency/emergency cases
• Admissions / discharges ratios • Occupancy rate • Census • % of urgent cases • Ratio of patients discharged by noon
- Understand if there are problems of scheduling and capacity planning - Analyze the horizontal interdependencies between production units - Identify sources of artificial variability - Identify bottlenecks
Surgical pipelines Medical pipelines Emergency pipelines
Glossary Beds Occupancy Rate = hospital inpatient days / (365*hospital beds) Length of stay = hospital inpatient days / total cases Bed Turns= total cases / number of beds (day hospitals are not included) % of urgent cases = unscheduled cases / total cases Census = average daily number of patients
pipeline) /
ork pro
4
wpd
1
2
3
4
5
c
% of “direct line” patients = (patients that start and end their patient journey within the same
Fig. 1. Structure of the analytical framew
.2. Hospital pipelines
If we consider the various potential patient journeysithin the hospital (from the patient’s arrival at the hos-ital to the final discharge), it is possible to identify fiveistinct physical pathways (which we call pipelines):
. emergency pipelines (physical pathways travelled bypatients who access the hospital through the EmergencyDepartment);
. surgical pipelines (referring to patients who undergo asurgical procedure);
. medical pipelines (referring to patients who only needmedical treatment and do not pass through the operat-ing room);
. day-surgery/day-hospital (referring to patients whoundergo minor surgical procedures or medical treat-ments and do not need a hospital bed);
. outpatient (referring to patients who stay within thehospital for a few hours during a medical or diagnostic
visit).
We identified these five mutually exclusive pipelinesrossing three different dimensions: (i) the type of
(total pipeline cases)
posed to measure patient flows logistics.
production units used (e.g., medical wards vs. surgicalwards), (ii) the type of demand met (particularly scheduledvs. unscheduled cases) and (iii) the patient’s length of staywithin the hospital (e.g., patients in the outpatient pipelinewho use hospital infrastructures for only a few hours).
It is important to stress that each pipeline passesthrough different possible production units (the third levelof this analysis) that can be exclusively assigned to a singlespecific pipeline (e.g. operating room or surgical beds in thecase of surgical pipeline) or can be shared between differentpipelines; for example a CT-scan can be used, alternatively,by surgical elective cases, outpatient cases and unsched-uled (emergency/urgency) patients.
Furthermore, it is, however, important not to confusethe production pipelines described in this paper with theclinical pathways [42]. The clinical pathways consist of theseries of physician and nursing activities performed withsub-groups of patients who are homogeneous under a clin-ical perspective, but not under a logistics perspective. Infact, clinical pathways related to complex conditions may
cross different hospital production pipelines. For exam-ple, a colon cancer patient who has been diagnosed duringan outpatient visit (outpatient pipeline) will perform thepre-operative screening in the Outpatient Department
Policy 1
across the different days. Furthermore, on average, somedays of the week (specifically Wednesdays) were charac-terized by higher percentages of urgent cases, suggesting
2 We are not referring here to actual clinical urgencies but to all thoseunscheduled patients that, for a variety of possible reasons, arrive at ahospital production unit without a previous appointment. This type of
200 S. Villa et al. / Health
(outpatient pipeline) before undergoing surgery (ordinarysurgical pipeline). Finally, after surgery, he/she might berequired to go through pharmaceutical treatment in a dayhospital (day hospital pipeline).
The analysis at the pipeline level is very useful to under-stand patient flow variability (across the day, the weekand the year) using typical variability indicators such as:standard deviation, range and variance coefficient. It isalso useful to analyze the transfers of patients betweenpipelines and to understand the interactions between elec-tive cases and urgent/emergency cases.
4.3. Production units
In the final step of the analysis, we analyze the capacity-utilization patterns of the various production units alongthe health care chain, including: operating rooms, emer-gency department, wards and ICUs (Intensive Care Units).This level of analysis is essential in understanding if thereare specific problems in the scheduling of specific produc-tion units; highlighting the horizontal interdependenciesbetween production units and anticipating any possibleproblems of coordination. In addition, it helped to identifyall the possible sources of artificial variability and the pro-duction unit that represents the bottleneck and that slowsdown the overall hospital throughput.
This three-layer framework considers all the differ-ent dimensions of patient flow logistics identified in thescientific literature. For example, the issue of capacityshortage [34,35] is captured at the hospital level, the anal-ysis of patient flow variability [10,11,34,36] is examined atpipeline level while problems of scheduling or bottleneckare caught at the level of single production unit [36,40,41].
It is important to highlight that the framework pre-sented in the paper is aimed to provide to managers andpolicy-makers a comprehensible and manageable tool tomeasure system-wide hospital patient flow logistics per-formance and, on the contrary, it does not provide solutionsto deal with operational level issues such as the schedulingof patients for appointments, or the organization of daysurgery procedures or the planning of surgical admissions.
5. Results
As already mentioned in Section 2, cases were chosenfor theoretical, and not statistical reasons. As summarizedin Table 1, hospitals included in the sample vary in differ-ent dimensions: size, type of ownership (private for profit,private not profit and public), teaching status and level ofspecialty (general vs. specialized). The characteristics of thehospitals included in the study are the results of a theo-retical sampling strategy in which we sought coverage oforganizational dimensions. The purpose of the study was,in fact, to develop and test a theoretical model capable ofanalyzing hospital-wide patient flow performance in anytype of hospital organization.
As illustrated in Fig. 1 our theoretical model is struc-
tured along three different levels: (i) hospital; (ii) pipelineand (iii) production units.
Some of the results of the analysis conducted at thehospital level are summarized in the two-by-two matrix
15 (2014) 196–205
in Fig. 2, in which two different dimensions are crossed:(i) bed-utilization rate and (ii) bed turns. Surprisingly, thehospitals with low occupancy rates and low levels of bedsturns (quadrant III of Fig. 2) were those with the mostcomplaints regarding issues such as delays, long waitingtimes and shortage of beds. It is, thus, clear that in all thesecases patient flow problems are not caused by a shortageof capacity, which is one of the possible scenarios indicatedin the scientific literature [5,34].
The analysis of the whole hospital system must be fol-lowed by a detailed analysis of each hospital pipeline. Inthe following sections, we present examples of the analysisconducted at the pipeline level, referring mainly to the caseof the surgical pipeline that is the most complex (becauseit involves different production units) and the most costly.
As suggested by several authors [1,7–11,34,38], a keyaspect that must be controlled, in the analysis of hospitalpipelines, is the variability of patient arrivals. This variabil-ity determines peaks and valleys in the demand for hospitalresources and creates queues, delays and stress for the hos-pital personnel. Therefore, it is important to identify andeliminate artificial variability.
Table 2 displays the variability of surgical cases in thesix hospitals included in the study. Particularly, tradi-tional variability indicators (such as standard deviationsand coefficients of variation) were computed for electiveand urgent/unscheduled cases2. The pattern exhibited bythese data, which is very similar among the six hospi-tals included in the comparative analysis, revealed twoimportant features. First, despite common beliefs, theurgent/unscheduled cases were less variable at aggregatelevel and much more predictable than the elective cases3.As confirmed by another study [10], the presumably con-trollable flow of patients scheduled to arrive for electiveprocedures is, in fact, more variable from day to day andweek to week than the unpredictable flow of patients beingadmitted due to emergencies, which, on the contrary, ismore predictable and stable over time.
If the final goal is to address the actual sources of artifi-cial variability, then it is necessary to extend the analysis toeach production unit. In the case of the surgical pipeline, acritical production unit that deeply influences the through-put time for the whole chain is represented by the operatingroom.
The hospital data of one hospital (Table 3) is a goodexample of this variability. In this case, the scheduling ofoperating rooms suffered from a variety of issues. First, theaverage utilization rate was modest (46% or 60% if we donot consider weekends), with significant peaks and valleys
information is included in hospital discharge data.3 This assumption also holds true when we do not consider holiday
weeks (e.g., Christmas, Easter or mid-August). In this case, when we talkabout urgent cases, we refer to aggregate levels of data (e.g. the numberof urgent cases per day).
S. Villa et al. / Health Policy 115 (2014) 196–205 201
Table 1Main characteristics of the hospitals included in the study.
Hospital Type of institution Structural data Clinical activity
Hospital 1 Public hospital belonging to a LocalHealth Authority
490 Inpatient beds 8 operatingrooms
Multi specialties hospital without cardiac surgerydepartment. 47,544 ED accesses per year
Average number 3 4 6 3 4 2 1 3Distribution % 12% 16% 28% 12% 20% 11% 1% 100%
Elecetive casesAverage number 37 40 38
Distribution % 19% 21% 19%
15
3
70%
75%
80%
85%
90%
95%
100%
35 40 45 50 55
Bed
s U
tiliz
atio
n R
ate
Beds
III
IV
Fig. 2. Patient flows analysis at hospital leve
34 38 8 0 2817% 20% 4% 0% 100%
2
46
60 65 70 75 80 Turns
I
II
l: beds utilization rate vs. bed turns.
202 S. Villa et al. / Health Policy 115 (2014) 196–205
Table 4Correlation between ED Length of Stay and ratio between discharges and admissions.
n. Observations Mean Standard deviation Correlation index ED LOS—Ratio OUT IN
Hospital 1 −0.19ED LOS Mean 187 24.26Ratio OUT/IN 1.00 0.44Hospital 2 335 −0.18ED LOS Mean 121.47 30.16Ratio OUT/IN 1.55 1.67Hospital 3 335 −0.15ED LOS Mean 91.59 16.35Ratio OUT/IN 1.05 0.39Hospital 6 335 −0.14ED LOS Mean 330 115Ratio OUT/IN 0.98 0.25
3
Hospital 5 335
ED LOS Mean 247
Ratio OUT/IN 1.15
that the urgent nature of these cases is somewhat ques-tionable. Lastly, the utilization rate dropped dramaticallyon specific days. In all these cases, we discovered that someof the clinical-specialty staff were away from the hospitaldue to conferences and/or training sessions.
The variability in the operating room utilization rate dis-played in Table 3 was common in all six hospitals includedin the study.
Lastly, other authors – particularly Vissers [34,35] –report that, in addition to the shortage of capacity and vari-ability, a third patient flow problem is the lack of coordina-tion between the different pipelines and production units.
Our data confirmed this hypothesis. First, we noticedthat the daily distribution of admissions and dischargeshad an impact on the ED length of stay (the amount oftime from the moment the patient arrives to the moment
the patient leaves). In fact, as presented by the correla-tion analysis in Table 4, there was (in all six hospitals)a negative correlation between the discharges/admissionsratio and the ED length of stay; that is, as the inpatient
Fig. 3. Distribution of ED arr
−0.161.480.62
discharges/admissions ratio increased, the following EDlength of stay consequently decreased.
This evidence is confirmed by other studies [11,43,44]showing that when beds on the floors are full, patientswho come in through the emergency department cannotbe admitted in a timely manner. Consequently, as deter-mined from the interviews, backups occur and patients areboarded in the hallways.
Second we analyzed the hourly distribution of admis-sions to the ED and the hourly distribution of discharges atthe only hospital for which we had this second set of data.As displayed in Fig. 3, there is a mismatch between the timeED patients are admitted and the time hospitalized patientsare discharged. Basically, the ED is a production unit thatfills up in the morning. Approximately 20% (in our popu-lation study) of these patients are kept in the hospital, so
they require a bed. In contrast, wards discharge the major-ity of their patients after noon, which creates a mismatchbetween supply and demand that causes daily fights forinpatient beds during the central hours of the day.
ivals and discharges.
Policy 1
aytwhdvgh
ailpotots
imw
ahftatcflefb
afstwdblfldFul
6
artiit
S. Villa et al. / Health
The insights derived from the graphic representationre also confirmed by the results of the correlation anal-sis. We found a strong (.4) negative correlation betweenhe ED length of stay and the daily percentage of patientsho were discharged by noon. This correlation was muchigher (.6) if we focused only on medical patients for twoifferent reasons: (i) the discharging process was moreariable for medical pipelines and (ii) compared with sur-ical departments, medical departments receive a muchigher percentage of ED cases (12% vs. 73%).
The use of the percentage of patients discharged by noons a possible explanatory variable of the ED length of stays definitely a novelty among the other studies [43,44] that,ike ours, have studied the relationships between ED hos-ital stays and the availability of inpatient beds. The hourf patient discharge provides practical hints. In fact, forhose treating patients in inpatient wards, the percentagef patients discharged by noon is a simple and intuitive toolhat can help guide physicians and nurses in finding newolutions to streamline the discharge process.
In conclusion, we can say that the test of the modeln the six hospitals included in the study reveals that the
odel is an appropriate tool to investigate the hospital-ide impacts of patient flow problems at the hospital level.
Particularly, the three-layer structure seems to be anppropriate way to frame a comprehensive appraisal ofospital patient flow performance for, at least, two dif-
erent sets of considerations. First, as already mentionedhroughout the paper, it is critical to adopt a system-widepproach to patient flow management and, thereafter, con-extualize any specific analysis within the overall hospitalontext (the first level of our analysis). Secondly, the dif-erent possible elements (as those found in the scientificiterature and those that emerged during the focus groups)xplaining hospital patient flow problems reside at dif-erent levels and are often due to a lack of coordinationetween pipeline and production units.
To this last regard, even if our findings cannot be gener-lized (given the small sample analyzed), the test of theramework in the six cases analyzed provides evidenceupporting some thesis advanced by other authors abouthe causes of hospital patient flow problems. Particularly,hile shortage of capacity does not seem to be a relevantriver, our data shows that patient flow variability causedy a bad allocation of capacity does represent a key prob-
em. Another aspect that is important to address in patientow change strategies is represented by the lack of coor-ination between different pipelines and production units.inally, the problem of overlapping between elective andnscheduled cases can be solved by setting aside a certain
evel of capacity to meet the unscheduled demand.
. Conclusions
The findings of the study offer also useful insightsnd food for thought for managers, policy-makers andesearchers. As for managers, the empirical test of the
heoretical model carried out among the six hospitalsncluded in the study highlights the importance of adopt-ng a system-wide approach capable of taking into accounthe interdependencies between different pipelines and
15 (2014) 196–205 203
production units. In this regard, managers who intendto lead patient flow change strategies should be awareof the importance of having a 360◦ view (the so-calledhelicopter view) achievable only thorough the creation ofa centralized unit/office in charge of patient flow logistics.
Second, to establish coherent and effective patient flowcontrol systems, it is very important to have real-time con-trol of the each patient’s entire hospital journey, from thevery first access (through the OPD or the ED) to the final dis-charge. The current configuration of hospital informationsystems is not capable of achieving this goal and thereforeneeds radical changes.
In fact the test of the model in the six hospitals revealedsome gaps in the current hospital information systemstructure. First, the databases built for each production unituse different patient identification codes; thus, it is verydifficult to keep track of the entire patient journey, espe-cially if patients cross various pipelines and productionunits. Second, there are often different interpretations ofthe same data – even within the same hospital – result-ing in serious problems with data validity. Third, somedata (for example, the ED length of stay or the numberof OR accesses) present serious problems of reliability.Fourthly, some data, useful for setting up effective patientflow control systems are, in most of the cases analyzed, notcollected, such as: (i) the hourly stays of patients in ICUs,(ii) the exact hour of patient discharge, and (iii) the internaltransfers between hospital pipelines and production units.
Third, the findings of this paper shed a light on somecommon misplaced beliefs emerged also during the focusgroups conducted in the study. For example, the data aboutthe low variability of emergency/urgency cases reveals thatthe often quoted problem of the overlapping between elec-tive cases and unscheduled cases can be solved by settingaside a certain production capacity dedicated to addressthis type of demand. Consequently, managerial attentionand efforts should be focused on identifying and eliminat-ing the sources of artificial variability.
Our study also offers useful insights for policy makers(regional governments, states or the European Union). First,to support the development of better patient flow man-agement systems at the hospital level, institutions shouldask hospitals (public and accredited by the national healthcare service) about homogeneous and standardized indica-tors of patient flow logistics. The framework presented inthis paper represents a useful starting point. The definitionof a minimum set of indicators is a necessary prerequisitefor the definition of the basic characteristics of the hos-pital information systems necessary to effectively supportpatient flow management programs. The development ofa common methodological framework to support analysisand changes in patient flow management is also a prerequi-site to carry out meaningful benchmarking initiatives bothat national and international level. Furthermore, the resultsof these benchmarking analyzes should favor the diffusionof standards and best practices contributing, in such a way,to strengthen a regional, national or even European identity
of hospital care.
Second, many of the patient flow problems identifiedin the study require infrastructural solutions and long-term investments that are not feasible without the support
Policy 1
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
204 S. Villa et al. / Health
of national or regional institutions. This consideration isalso confirmed by other studies [3,6,7,9,13] that show howmany innovations in patient flow logistics have been madepossible thanks to the strong external pressures posed byinstitutions. In this sense, policy-makers can play a verycritical role in supporting change and in diffusing best prac-tices in patient flow management.
In the specific case of Italy, given the massive process ofdecentralization occurred from the beginning of 2000, theresults of this study are particularly useful for regional gov-ernments. The Regions now have in fact jurisdiction over allhealth care issues and have started to reshape their systemsbased on different ideas and approaches to health care poli-cies [45]. In this scenario, Regions are also required to leadand promote innovation in the organization and deliveryof care in the hospitals included in the system.
For researchers, finally, the study offers interesting con-siderations to reshape future research in this field. It isimportant to adopt a system-wide approach in the def-inition of the research design. Even operations researchand management studies that focus on single sections ofa hospital (e.g., a single hospital floor or the operatingroom) should always anticipate the possible impact of thesuggested solutions on the other hospital pipelines andproduction units. If these studies fail to analyze and explainthese interdependencies, they run the risk of providingbiased assessments and, consequently suggesting faultysolutions.
In addition, as already mentioned, the current hospitalinformation system needs to be redesigned to better sup-port patient flow management strategies. Thus, there is aneed for robust and evidence-based studies capable of indi-cating standards and guidelines, regarding the proceduresto be followed for data collection and the construction ofindicators to set up effective patient flow control systems.The creation of a standardized flow of data and informa-tion is an essential prerequisite for establishing effectiveand consistent policies on patient flow management at theinstitutional level.
References
[1] Haraden C, Resar R. Patient flow in hospitals: understanding andcontrolling it better. Frontiers of Health Services Management2004;20(4):3–15.
[2] Vissers J, Beech R. Health operations management. Routledge: Oxon;2005.
[3] Villa S, Barbieri M, Lega F. Restructuring patient flow logistics aroundpatient care needs: implications and practicalities from three criticalcases. Health Care Management Science 2009;12:155–65.
[4] Butler TW. Management science/operations research projects inhealth care: the administrator’s perspective. Health Care Manage-ment Review 1995;20(1):19–25.
[5] De Vries G, Bertrand J, Vissers J. Design requirements for healthcare production control systems. Production, Planning & Control1999;10(6):559–69.
[6] Davies C, Walley P. Clinical governance and operations managementmethodologies. International Journal of Health Care Quality Assur-ance 2000;13(1):21–6.
[7] Noon CE, Hankins CT, Côté MJ. Understanding the impact of vari-ation in the delivery of healthcare services. Journal of HealthcareManagement 2003;48(2):82–97.
[8] McCarthy M. Can car manufacturing techniques reform health care?The Lancet 2006;367(9507):290–1.
[9] Walley P, Steyn R. Managing variation in demand: lessons fromthe UK National Health Service. Journal of Healthcare Management2006;51(5):309–20.
[
15 (2014) 196–205
10] Litvak E, Buerhaus P, Davidoff F, Long M, McManus M, Berwick D.Managing unnecessary variability in patient demand to reduce nurs-ing stress and improve patient safety. Journal on Quality and PatientSafety 2005;31(6):330–8.
11] Litvak E, Bisognano M. More patients, less payment: increasing hos-pital efficiency in the aftermath of health reform. Health Affairs2011;30(1):76–80.
12] Mintzberg H. Managing care and cure—up and down, in and out.Health Service Management Research 2002;15(3):193–206.
13] Lega F, De Pietro C. Converging patterns in hospital organization:beyond the professional bureaucracy. Health Policy 2005;74:261–81.
14] Radnor ZJ, Holweg H, Waring J. Lean in healthcare: the unfilledpromise? Social Science and Medicine 2011;74(3):364–71.
15] Mazzocato P, Savage C, Brommels M, Aronsson H, Thor J. Lean think-ing in healthcare: a realist review of the literature. Quality & Safetyin Health Care 2010;19(5):376–82.
16] Restuccia J. Transformational change in health care: an organiza-tional model. Health Care Management Review 2007;32(4):309–20.
17] Eisenhardt KM. Building theories from case study research. Academyof Management Review 1989;14(4):532–50.
18] Yin RK. Case study research: design and methods, 2nd ed. BeverlyHills, CA: SAGE Publications; 1994.
19] Miles MB, Huberman AM. Qualitative data analysis, 2nd ed. NewburyPark, CA: Sage; 1994.
20] Edmondson AC, Mcmanus SE. Methodological fit in manage-ment field research. Academy of Management Review 2007;32(4):1155–79.
21] Bradfield R, Wright G, Burt G, Van der Heijden K. The origins andevolution of scenario techniques in long range business planning.Futures 2005;37(8):795–812.
22] Maruster L, Weijters T, De Vries G, Van Den Bosch A, DaelemansW. Logistic-based patient grouping for multi-disciplinary treatment.Artificial Intelligence in Medicine 2002;26:87–107.
23] Ravn H, Petersen LO. Balancing the surgical capacity in a hospital.International Journal of Healthcare Technology and Management2007;8(6):603–24.
24] Van Sambeek JRC, Cornelissen FA, Bakker PJM, Krabbendam JJ.Models as instruments for optimizing hospital processes: a system-atic review. International Journal of Health Care Quality Assurance2008;23(4):356–77.
25] Katharaki M. Approaching the management of hospital units withan operation research technique: the case of 32 Greek obstetric andgynaecology public units. Health Policy 2008;85:19–31.
26] Lane DC, Husemann E. System dynamics mapping of acute patientflows. Journal of the Operational Research Society 2008;59(2):213–24.
27] Litvak N, Van Rijsbergen M, Boucherie RJ, Van Houdenhoven M. Man-aging the overflow of intensive care patients. European Journal ofOperational Research 2008;185(3):998–1010.
28] Hanne T, Melo T, Nickel S. Bringing robustness to patient flowmanagement through optimized patient transports in hospitals.Interfaces 2009;39(3):241–55.
29] Zonderland ME, Boucherie RJ, Litvak N, Vleggeert-Lankamp CLAM.Planning and scheduling of semi-urgent surgeries. Health Care Man-agement Science 2010;13:256–67.
30] Conforti D, Guerriero F, Guido R, Cerinic MM, Conforti ML. An optimaldecision making model for supporting week hospital management.Health Care Management Science 2011;14:74–88.
31] Sciomachen A, Tanfani E, Testi A. Simulation models for optimalschedules of operating theatres. International Journal of Simulation2005;6(12/13):26–34.
32] Joustra PE, De Wit J, Van Dijk NM, Bakker PJM. How to juggle priori-ties? An interactive tool to provide quantitative support for strategicpatient-mix decisions: on ophthalmology case. Health Care Manage-ment Science 2011;14(4):348–60.
33] Borgonovi E, Zangrandi A. L’ospedale. Un approccio economico azien-dale. Milano: Giuffrè; 1998.
34] Vissers J, De Vries G. A framework for production controlin health care organizations. Production, Planning and Control2001;12(6):591–604.
35] Vissers J. Patient flow-based allocation of inpatient resources: a casestudy. European Journal of Operational Research 1998;105:356–70.
36] Litvak E, Long MC. Cost and quality under managed care: irrec-oncilable differences? The American Journal of Managed Care
2000;6(3):305–12.
37] Chand S, Moskowitz H, Norris JB, Shade S, Willis DR. Improvingpatient flow at an outpatient clinic: study of sources of variabil-ity and improvement factors. Health Care Management Science2008;12:325–40.
38] Bekker R, Koeleman PM. Scheduling admission and reducingvariability in bed demand. Health Care Management Science2011;14(3):237–49.
39] Villa S. L’operations management a supporto del sistema di oper-azioni aziendali. Modelli di analisi e soluzioni progettuali per ilsettore sanitario. Padova: CEDAM; 2012.
40] Young T, Brailsford S, Connell C, Davies R, Harper P, Klein JH. Usingindustrial processes to improve patient care. British Medical Journal
2004;328(7432):162–4.
41] McLaughlin DB, Hays JM. Healthcare operations management.Chicago: Health Administration Press; 2008.
42] Baraghini G, Capponi A, Capelli M, Longo F, Tozzi V, Villa S. La ges-tione per processi per il governo clinico e il controllo dei rischi: un
[
15 (2014) 196–205 205
confronto di metodo tra Aziende Ospedaliere. In: Anessi Pessina E,Cantù E, editors. Rapporto OASI 2006. L’aziendalizzazione della sanitàin Italia. Milano: EGEA; 2006. p. 495–530.
43] Forster AJ, Stiell I, Wells G, Lee AJ, Van Warlaven C. The effect of hos-pital occupancy on emergency department length of stay and patientdisposition. Academic Emergency Medicine 2003;10(2):127–33.
44] Vermeulen MJ, Ray J, Bell C, Cayen B, Stuckel TA, Schull MJ. Dis-equilibrium between admitted and discharged hospitalized patients
affects emergency department length of stay. Annals of EmergencyMedicine 2009;54(6):794–804.
45] Anessi Pessina E, Cantù E, Jommi C. New funding arrangements inthe Italian National Health Service. International Journal of HealthPlanning and Management 2001;16(4):47–68.
