The study was conducted in 2 parts. A prospective, descriptive, longitudinal study was performed to analyze real-world data from a group of patients with diabetes with infected foot ulcers; the known prevalence of microorganisms and registered treatment costs were described. In the second step, a hypothetical cohort with 100 iterations (case simulations) at different treatment stages was constructed through the Monte Carlo simulation, employing the data derived from the patients studied in the first phase (this information was utilized for the mathematical basis of the economical tool). The simulations provided predictions of the economic impact of obtaining early microbiological cultures of infected ulcers in order to make an informed decision about a specific antibiotic treatment.
The study was approved by the CLIEIS No. 1306 (Comité Local de Investigación y Ética en Investigación en Salud: Local Committee of Research and Ethics in Health Research) at the Institute Mexicano del Seguro Social (IMSS) Jalisco. Based on the observational nature of the investigation, the Committee authorized the use of signed informed consent according to the guidelines in the regulation of the General Health Law responsible for health research of studies without risk in Mexico.
In this study, data were obtained from the medical charts of patients with diabetes with IDFU who were seen at the IMSS Emergency Room (ER) of a second-level facility from January 1 to April 30, 2010. The patients were followed during hospital treatment for the IDFU. Patients with a diagnosis of cancer, hepatic cirrhosis, and/or rheumatic-orthopedic diseases that could impact their walking ability and patients with pressure-associated ulcers or DFU that appeared during their hospital stay were excluded.
Cost per patient was calculated based on the identification of the antibiotic schedule, the bacterium, and antimicrobial resistance from the patients’ charts. Length of hospital stay (days); number and type of laboratory tests, medications, and intravenous (IV) fluids; costs for surgical and supportive treatment; specialists (type and number of); and clinical outcomes also were abstracted, as well as demographic information and the patient’s history (age, gender, marital status, and time elapsed since first DM2 diagnosis). The Wagner classification of foot ulcers and microbiological culture also were obtained. The cost of each component of the medical treatment was obtained from the IMSS website IMSS compró21 (an institutional website that displays for public opinion the cost of drugs, medical technology, wound care material, and other supplies purchased by IMSS at each hospital).
Data collection. Data were collected from the patient’s chart by an emergency medicine physician and subsequently entered into an electronic database. To preserve anonymity, patients were identified by means of a code and their social security number.
Statistical analysis. Using the data collected by Gutiérrez et al,3 the sample size of this study was calculated with a 95% confidence interval (95% CI) and 80% potency. For quantitative data, means, standard deviations (SD), medians, ranks, and intervals were calculated. For qualitative data, percentages and proportions were utilized. All data were processed in SPSS version 16 and Excel 2010 statistical software (SPSS Inc, Chicago, IL).
In parallel, a descriptive economic analysis using microcosting was performed to evaluate the partial cost of medical treatment by employing the third-party-payer perspective (ie, the IMSS, for this particular study). The time horizon defined was 30 days. Discount rates were not applied.
Microcosting comprises a cost estimation method that involves the “direct enumeration and costing out of every input consumed in the treatment of a particular patient.”21 It employs the aggregation level, which allows knowing the lower-level estimated cost associated with the various treatment elements for a given level within the patient’s treatment. Indirect costs such as buildings, equipment, and maintenance were not considered. In order to estimate unit costs, the resource pattern utilized was identified according to microorganism type and was registered in the medical record of each patient; a monetary value was assigned later. The unit cost was identified for each resource. For health services (consultation, surgeries, hospitalization), cost information was obtained from the Diario Oficial de la Federación (DOF) website22 and the medication cost was obtained from the IMSS website under IMSS compró.23 The economic expenditure for each resource was calculated as the product of the unit cost multiplied by the number of units used. The monetary value for Mexican pesos was obtained in 2009 and converted into US dollars (USD) with an exchange rate of $12.75 Mexican pesos per $1 USD.
The following cost-related information was calculated: total cost per patient, total cost for the group of patients with diabetes, and mean cost for each patient.
Markov modeling. This mathematical model is employed in decision analysis to evaluate potential outcomes of a disease process, which are defined as specific health states and transitions that are modeled iteratively. In standard-decision tree analysis, a patient moves through states in a Markov process. Some states cannot be left once entered (the so-called “absorbing states” or natural history of the disease), including death.24-26 In IDFU, specific issues were considered with the following elements in the process: DM2 without complications; DM2+DFU; DM2+DFU+infection; DM2+DFU+infection+1st amputation; and DM2+DFU+1st amputation+2nd amputation (see Figure 1 and Tables 1 and 2), to estimate (calculated values) the proportion of persons affected by each condition per specific period (eg, proportion of patients with DM2+DFU in the first period [0.25], which corresponds to 25%; while in the second period, 0.297 corresponds to 29.7%). These proportions are utilized to construct the probabilistic mathematical model.
Monte Carlo modeling. The Monte Carlo method is a nondeterministic statistical method for obtaining a numerical solution for a problem too complicated to be solved analytically. This method solves a problem by generating a set of random or pseudorandom numbers within the domain of the variables under study. The corresponding absolute error decreases as the number of Monte Carlo evaluations increases, using the central limit theorem as a basis.27 The mathematical model allows for containing estimated numeric results through a design of specific clinical sets.28
The Monte Carlo predictive model was used in this study to find the behavior (through extrapolating the data of a specific clinical set) of the phenomenon denoted infected diabetic foot ulcer. Differences between real cost and estimated cost of procedures including early microbiological culture in each patient (eg, surgical procedures, wound healing, culture specimens) were considered based on the statistical results to optimize resources for the patient and for the third-party payer (IMSS) (see Figure 2 and Table 3). The Monte Carlo simulation was employed to estimate the cost reduction associated with early identification of the specific microorganism in the IDFU at the beginning of treatment. Data for the model included species, type of bacterium according to Gram stain, number of infecting microorganisms, and antibiotic sensitivity through bacterial culture.
The information analyzed included the bacterium type identified through microbiological culture; number of microorganisms identified in each microbiological culture and their susceptibility to medications (1–3 different bacteria were identified); unit cost for each medication (purchase of ordinary [eg, public tendering — ie, an administrative procedure where governments select the person or vendor from whom to buy different kind of supplies through public contest; and centralized procurement coordinated through the Acquisitions Department] and extraordinary [eg, items purchased from suppliers without public tendering such as antibiotics) used in the treatment recommended by the physician; length of ER stay (days), mindful of guidelines that have established a maximum time in the ER as 24 hours, while every extra hour generates an increase in financial expenditures per patient; length of hospital stay (days); and cost per day for total hospital stay at a secondary-level hospital.
The structure of the Monte Carlo model involved medications (antimicrobial spectrum: active against Gram-positive bacteria, Gram-negative bacteria, or anaerobic bacteria); specific antibiotic; cost of each antibiotic; cost of each group of antibiotics; and bacteria (specific bacterium identified through microbiological culture, along with bacterial group [Gram-positive, Gram-negative, anaerobic bacteria, fungus]).
As part of the Monte Carlo model, the following factors were considered: days of hospital stay (total or overnight and in the ER); hospital-associated costs; type of bacterium or group of bacteria (see Table 4); specific antibiotic or antibiotic groups, and the cost of each antibiotic in ordinary and extraordinary purchasing (see Table 5). One hundred (100) Monte Carlo iterations (simulations) were performed in order to observe the predictive model behavior. Final treatment cost and total cost with the proposed model were ultimately analyzed.