Poching DeLaurentis

Using no-show modeling to improve clinic performance.

‘No-shows’ or missed appointments result in under-utilized clinic capacity. We develop a logistic regression model using electronic medical records to estimate patients’ no-show probabilities and illustrate the use of the estimates in creating clinic schedules that maximize clinic capacity utilization while maintaining small patient waiting times and clinic overtime costs. This study used information on scheduled outpatient appointments collected over a three-year period at a Veterans Affairs medical center. The call-in process for 400 clinic days was simulated and for each day two schedules were created: the traditional method that assigned one patient per appointment slot, and the proposed method that scheduled patients according to their no-show probability to balance patient waiting, overtime and revenue. Combining patient no-show models with advanced scheduling methods would allow more patients to be seen a day while improving clinic efficiency. Clinics should consider the benefits of implementing scheduling software that includes these methods relative to the cost of no-shows.

Hospital stockpiling for disaster planning

In response to the increasing threat of terrorist attacks and natural disasters, governmental and private organizations worldwide have invested significant resources in disaster planning activities. This article addresses joint inventory stockpiling of medical supplies for groups of hospitals prior to a disaster. Specifically, the problem of determining the stockpile quantity of a medical item at several hospitals is considered. It is assumed that demand is uncertain and driven by the characteristics of a variety of disaster scenarios. Furthermore, it is assumed that hospitals have mutual aid agreements for inventory sharing in the event of a disaster. Each hospitals desire to minimize its stockpiling cost together with the potential to borrow from other stockpiles creates individual incentives well represented in a game-theoretic framework. This problem is modeled as a non-cooperative strategic game, the existence of a Nash equilibrium is proved, and the equilibrium solutions are analyzed. A centralized model of stockpile decision making where a central decision maker optimizes the entire system is also examined and the solutions obtained using this model are compared to those of the decentralized (game) model. The comparison provides some managerial insights and public health policy implications valuable for disaster planning.

Planning for Pandemic Influenza: Lessons from the Experiences of Thirteen Indiana Counties

Significant concerns exist over the ability of the healthcare and public health systems to meet the surge demands that would result from an event such as an influenza pandemic. Current guidance for public health planners is largely based on expert opinion and may lack connection to the problems of street-level public health practice. To identify the problems of local planners and prepare a state-level planning template for increasing health care surge capacity that accounted for these issues, a study was conducted of local pandemic planning efforts in thirteen counties, finding that cognitive biases, coordination problems, institutional structures in the healthcare system, and resource shortfalls are significant barriers to preparing and implementing a surge capacity plan. In addition, local planners identify patient demand management through triage and education efforts as a viable means of ensuring adequate capacity, in contrast to guidance proposing an increased supply of care as a primary tool.

Hospital stockpiling for influenza pandemics with pre-determined response levels

This paper explores the problem of hospital stock-piling of critical medical supplies in preparation for a possible influenza pandemic.We consider a regional network of hospitals that have mutual aid agreements in place such that they may borrow or lend supplies from each other during medical emergencies. We assume that the attack rate is a random variable with known distribution and that the demand surge due to the pandemic is a function of the attack rate, and is thus stochastic. We further assume that each hospital in the network pre-determines its targeted pandemic response level, and that any demand beyond this pre-determined level is reallocated to other hospitals. Each hospital in the network must decide the stockpile level that minimizes its expected overall cost, including purchasing cost, holding cost, cost (revenue) for borrowing (lending), penalty for setting a too low targeted level, and shortage penalty. To capture the mutual aid relationships of hospitals in the network, we formulate the problem as a game theoretic model. We show that the response sets are nested and we provide an algorithm to obtain numerically the Nash Equilibrium solution of this game. We illustrate the structure of the model on a two-hospital example and perform sensitivity analysis with respect to parameters of our model.

Infusion pump informatics approach to quantify impact of alerts and alarms on healthcare delivery

Infusion pumps are medical devices that deliver medication, fluids, and nutrients in a precise and controlled manner that is critical to patient care. This study proposes using infusion pump informatics on all-infusion datasets to understand current impact of alerts and alarms on patient care and health practitioner workflow. All-infusion datasets contain infusion data for both normal and abnormal use, i.e., error states. Ten months of continuous data was collected from one health institution. Analysis of variance with log-transformation and logistic regressions were used to analysis contributing factors for alerts and alarms states. A total 64,511 minutes of alarm activation were observed, where alarms were active prior to being resolved. Mean resolution times for 83% of alarms were one minute or less; however, 3% or alarms required >4 minutes before getting resolved. Risk factors for infusions with alerts included nursing shift variables. Specifically, odds for alerts were 1.3 times higher for infusions that span across shifts than infusions in the day shift.

Consideration of system of systems and service systems as complimentary approaches for healthcare problems

The purpose of this paper is to consider key features of a system of systems and service systems approach in the context of healthcare service. The investigation is motivated by the increasing demand for improvement in patient outcomes in terms of quality, reliability and cost of care. We hypothesize that an integration of system of systems and service systems approaches may yield benefits in terms of both correct and complete representation of healthcare services and applicability of associated modeling tools to solve problems. The paper shows an initial mapping of this integration after the salient features of the two bodies of knowledge are described in the context of an information intensive healthcare delivery system.

Prevalence of wireless smart-pump drug library update delays

Purpose. Results of a study to estimate the prevalence and severity of delays in wireless updates of smart‐pump drug libraries across a large group of U.S. hospitals are reported. Methods. A prolonged smart‐pump drug library update may result in patient harm if a pump is programmed with an incorrect limit setting at the time of drug administration. A retrospective study was conducted using smart‐pump alert data extracted from the Regenstrief National Center for Medical Device Informatics (REMEDI) database. The study sample consisted of 49 hospitals in 5 states across the Midwest and Kentucky operated by 12 health systems; all the facilities used a specific brand of smart pump (BD Alaris, Beckton, Dickinson and Company) capable of generating alert data and had consistently contributed alert data to the REMEDI database over a 2‐year period. An update delay was defined as the interval from the time a drug library version was replaced to the time of the last infusion alert triggered by the previous version during the study period. Results. Of the 12 health systems, 11 were found to have had drug library update delays during the study period, with delay medians ranging from 22 to 192 days. The overall delay minimum and maximum durations were 0 and 661 days. Conclusion. Substantial delays in completion of wireless updates of smart‐pump drug libraries were common across a group of hospitals of various sizes.