Mathias Barra

Towards a multimethodology in health care – synergies between Soft Systems Methodology and Discrete Event Simulation

Multimethodology is the combination of methodologies, often from different paradigms. While methodologies from hard paradigms are rather positivist and treats the organizational world as objective, methodologies from soft paradigms are interpretivist by nature. Some argue that these paradigms are incommensurable and multimethodology is therefore a challenging research field. However, we believe this is possible within a pragmatic paradigm. In this article, a multimethodology combining Soft Systems Methodology (SSM) and Discrete Event Simulation (DES) is developed and tested at a Norwegian Hospital. The initial phases of the multimethodology anchored the project with the staff by basing the DES-model on their narratives. This phase also revealed real-world mechanisms vital to the DES-modelling. During later SSM-steps, the DES-model was used to detect possible areas of improvement. Our findings were well received by the hospital. Our work shows how soft and hard methodologies yield synergies when they are carefully brought together in a pragmatic fashion.

Pre-contact by telephone increases response rates to postal questionnaires in a population of stroke patients: an open ended randomized controlled trial

BackgroundA follow-up study on a cohort of stroke patients through a postal survey questionnaire 3 and 12 months after discharge from hospital was performed. The response rate at 3-months follow-up was lower than desired, and pre-contact by phone as a measure for increasing the response rate at 12 months was studied.MethodsThe study design was a randomized controlled trial on a cohort of 3 months follow-up-non-responders where the intervention group was pre-contacted with an aim to obtain an informal ’consent to receive’ the questionnaire before the 12-months survey was mailed, and the control group was not.The primary outcome was 45 days response rate; secondary outcome was 365 days response rate. The main analysis followed the intention to treat principle. A secondary, per-protocol analysis (i.e. subjects who were not reached by phone were reassigned to the control group) is included. Also included is a rudimentary cost-utility analysis, where we estimated the cost per additional response.ResultsOf the 235 subjects, 116 were randomized to the intervention group and 119 to the control group. 10 were excluded due to death (7 in the IG and 3 in the CG), 6 due to dementia (3 in the IG and 3 in the CG), and 2 (1 in the IG and 1 in the CG) for other reasons. The primary outcome was a response rate of 42.9 % in the intervention group, and 26.8 % in the control group, giving p =0.014, with estimated OR of 2.04 (95 % CI [1.16,3.64]). The secondary outcome had p =0.009 with OR 2.10 (95 % CI [1.20,3.70]). The as-per-protocol analyses gave stronger results with p =0.001 and p =0.003, respectively. The cost-utility analysis gave a time cost of 1 working hour per additional response.ConclusionsThe results are in line with previous research, and show that pre-contact has a positive effect on response rate also for a population of elderly with reduced health. Given the importance of high response rate in surveys, a cost of 1 working hour per additional response is likely to be worth while.Trial registrationRegistration with ISRCTN initiated on 05/21/2013 and finalised on 06/30/2014 with http://www.isrctn.com/ISRCTN31304930. Following the prospective submission in May 2013, there were no subsequent changes to the protocol. The recruitment started on 01/06/13, after initiation of public registration.

The consequences of how subject matter expert estimates are interpreted and modelled, demonstrated by an emergency department des model comparing triangular and beta distributions

The aim of this paper is to demonstrate empirically the consequences of misinterpreting estimates from subject matter experts (SMEs), and to study the differences between modeling this with triangular and beta distributions. Three estimates which describe the duration of a process; minimum, maximum, and mode, is ideally sufficient as a proxy for the empirical distribution. However, these estimates might be biased when the SMEs confuse the difference between mean and mode. The analysis are performed in an ED model of a Norwegian hospital. When comparing the model output with data from the electronic patient record we see that a model with beta distributions based on the SME estimates outperforms a model with the more frequently used triangular distributions. A triangular distribution will overestimate the mean of the distribution compared to a beta distribution. We therefore encourage the use of beta distributions over triangular for activities with skewed distributions.

The Burden of Stroke Mimics: Present and Future Projections

BACKGROUND AND PURPOSE An increasing proportion of patients presenting with suspected stroke prove to have other conditions, often referred to as stroke mimics. The aim of this study was to present a projection of the number of hospitalized strokes, transient ischemic attacks (TIAs), and stroke mimics in Norway up to the year 2050 based on expected demographic changes, to estimate the burden of stroke mimics in the coming decades. MATERIALS AND METHODS This prospective study included all admissions to the stroke unit of Akershus University Hospital from March 1, 2012, to February 28, 2013. Relevant resource use was recorded. Based on the age- and sex-specific absolute incidences for the study period, the expected numbers of strokes, TIAs, and stroke mimics in the entire Norwegian population were computed for every fifth year for the period 2020-2050. RESULTS We included 1881 admissions, of which 38.2% were stroke mimics. With constant age- and sex-dependent incidence rates, we estimated that the number of strokes and stroke mimics will respectively increase by 121.3% and 88.7% (men) and 97.6% and 71.7% (women). For hospital admission levels to stay constant at the 2013 level, an annual reduction of 2.1% and 1.7% (men) and 1.8% and 1.5% (women) must take place for strokes and mimics, respectively. CONCLUSIONS A significant proportion of stroke unit admissions prove to have other conditions than stroke. With constant age- and sex-dependent incidence rates, the number of stroke mimics admissions will increase substantially over the next decades.

The disvalue of death in the global burden of disease

In the Global Burden of Disease study, disease burden is measured as disability-adjusted life years (DALYs). The paramount assumption of the DALY is that it makes sense to aggregate years lived with disability (YLDs) and years of life lost (YLLs). However, this is not smooth sailing. Whereas morbidity (YLD) is something that happens to an individual, loss of life itself (YLL) occurs when that individual’s life has ended. YLLs quantify something that involves no experience and does not take place among living individuals. This casts doubt on whether the YLL is an individual burden at all. If not, then YLDs and YLLs are incommensurable. There are at least three responses to this problem, only one of which is tenable: a counterfactual account of harm. Taking this strategy necessitates a re-examination of how we count YLLs, particularly at the beginning of life.

Conflicts or synergy when combining modeling approaches: perspectives from psychology

The simulation modeling community consists of several frameworks or approaches that have been developed at different times to handle different problems, and persist in a state of relatively limited interaction. Various forms of hybrid modeling, combining aspects of two or more modeling approaches, have been proposed and used, but these still represent a relatively small part of the world of simulation modeling. In this paper, we will draw on parallels between the current debate around discrete event simulation and agent-based modeling, and the historic conflict between two schools of psychology: behaviorism (human thought considered a “black box”, focus restricted to observable behavior), and cognitive psychology (emphasis on conscious thought processes). Through a presentation of different perspectives on what happened in psychology, we will discuss views on the combination of different modeling approaches, and implications of similar perspectives on the future development of simulation modeling.

Do hybrid simulation models always increase flexibility to handle parametric and structural changes

Are hybrid simulation models always beneficial? When should one modeling paradigm be used more than another? How does one know the right balance has been reached between different simulation techniques for the system under investigation? We illustrate selected insights into hybrid simulation through the use of a discrete event simulation (DES) model and a hybrid DES agent based model (ABM) of the obstetrics department at Akershus University Hospital. Design decisions are not straightforward, and have different impacts on model development and ability to address different scenarios or potential changes. In the DES model, the majority of the logic and code representing patient pathways is contained within the structure of the model. In the AB-DES model, a selection of the code is shifted from the model structure to the patient entities. Scenarios are presented which illustrate strengths and weaknesses of each model. These are reflected on and future work is suggested.

Differences in and Determinants of Prehospital Delay Times among Stroke Patients—1994 Versus 2012

OBJECTIVES Prehospital delay is a challenge for stroke treatment and the delivery of time-critical treatments. Few studies have examined secular trends in prehospital delay, and results vary. This study investigates how prehospital delay among Norwegian stroke patients has changed over the last 2 decades. METHODS We compared time from symptom onset to admission in 2 cohorts of stroke patients admitted to Akershus University Hospital, Norway, in 1994 (n = 550) and 2012 (n = 522), and constructed predictive models for arrival within 3 hours for each cohort. RESULTS More patients arrived within 3 hours of symptom onset in 2012 compared to 1994 (proportion, 47.1% versus 19.3%, P < .001), also after adjusting for age, sex, and baseline differences; odds ratio (OR) was 5.14 (95% confidence interval [CI] 3.69-7.15). Stroke severity was the only predictor examined that was independently associated with early arrival during both periods. For patients with moderate strokes the overall OR was 2.06 (95% CI 1.41-3.00) and for severe strokes 4.52 (95% CI 2.97-6.87), compared to those with mild strokes. In the 1994 cohort additional predictors of early arrival were living with others and not being admitted from nursing home. CONCLUSIONS Prehospital delay in Norway has decreased considerably over the last 2 decades and since the availability of time-critical treatments. However, there is still an urgent need to reduce the number of delayed admissions as a large proportion of patients continue to arrive too late to benefit from these treatments. Patients with severer strokes were predicted to have earlier arrival.

Seasonally adjusted birth frequencies follow the Poisson distribution.

BACKGROUND Variations in birth frequencies have an impact on activity planning in maternity wards. Previous studies of this phenomenon have commonly included elective births. A Danish study of spontaneous births found that birth frequencies were well modelled by a Poisson process. Somewhat unexpectedly, there were also weekly variations in the frequency of spontaneous births. Another study claimed that birth frequencies follow the Benford distribution. Our objective was to test these results. MATERIAL AND METHOD We analysed 50,017 spontaneous births at Akershus University Hospital in the period 1999-2014. To investigate the Poisson distribution of these births, we plotted their variance over a sliding average. We specified various Poisson regression models, with the number of births on a given day as the outcome variable. The explanatory variables included various combinations of years, months, days of the week and the digit sum of the date. RESULTS The relationship between the variance and the average fits well with an underlying Poisson process. A Benford distribution was disproved by a goodness-of-fit test (p < 0.01). The fundamental model with year and month as explanatory variables is significantly improved (p < 0.001) by adding day of the week as an explanatory variable. Altogether 7.5% more children are born on Tuesdays than on Sundays. The digit sum of the date is non-significant as an explanatory variable (p = 0.23), nor does it increase the explained variance. INERPRETATION: Spontaneous births are well modelled by a time-dependent Poisson process when monthly and day-of-the-week variation is included. The frequency is highest in summer towards June and July, Friday and Tuesday stand out as particularly busy days, and the activity level is at its lowest during weekends.

Examining the relationship between health-related quality of life and increasing numbers of diagnoses

PurposeLittle is known about estimating utilities for comorbid (or ‘joint’) health states. Several joint health state prediction models have been suggested (for example, additive, multiplicative, best-of-pair, worst-of-pair, etc.), but no general consensus has been reached. The purpose of the study is to explore the relationship between health-related quality of life (HRQoL) and increasing numbers of diagnoses.MethodsWe analyzed a large dataset containing respondents’ ICD-9 diagnoses and preference-based HRQoL (EQ-5D and SF-6D). Data were stratified by the number of diagnoses, and mean HRQoL values were estimated. Several adjustments, accounting for the respondents’ age, sex, and the severity of the diagnoses, were carried out. Our analysis fitted additive and multiplicative models to the data and assessed model fit using multiple standard model selection methods.ResultsA total of 39,817 respondents were included in the analyses. Average HRQoL values were represented well by both linear and multiplicative models. Although results across all analyses were similar, adjusting for severity of diagnoses, age, and sex strengthened the linear model’s performance measures relative to the multiplicative model. Adjusted R2 values were above 0.99 for all analyses (i.e., all adjusted analyses, for both HRQoL instruments), indicating a robust result.ConclusionsAdditive and multiplicative models perform equally well within our analyses. A practical implication of our findings, based on the presumption that a linear model is simpler than an additive model, is that an additive model should be preferred unless there is compelling evidence to the contrary.