Antimicrobial control is imperative due to emergence of resistant isolates, shortage of therapeutic drugs, risk of clinical failure and mortality, and expenditure increases1,2. Recently, a global program has been issued by WHO asking to adopt different strategies to control this threat3. Effective interventions for hospitals have been delineated and the impact on consumption, expenditure, and sometime antibiotic resistance has been supported by systematic reviews4,5. These interventions require financial investment to hire human resources, and support differente activities. Chile has a governmental regulation issued by the Ministry of Health since 1999 in order to implement antimicrobial stewardship programs (ASP) in hospitals that has never been supervised, and studies developed more a decade ago showed partial fulfillment6. Literature on the impact of ASP in Chile is scarce7,8 and probably explained by the low number and the geographically concentrated distribution of infectious diseases (ID) specialists. Novel information on the impact of ASP on consumption, cost-saving, antimicrobial resistance, and mortality, is needed to convince hospital administrators to invest in dedicated human resources. In addition, most of the evidence available refers to short-term impacts, usually for less than 2 years5 with few reports measuring the effect beyond this frametime9. On ther hand, a progressive trend for competitive biddings with prequalified pharmaceutical providers, has been adopted in different hospitals, countries or worldwide programs allowing cheapest drug prices10,11. This policy may influence cost saving and its contribution should be measured when ASP are also implemented.
We aimed to evaluate the long-term impact of an ASP in a general hospital in Chile, its effect on consumption, mortality, expenditure and antimicrobial resistance. Besides, as competitive biddings occurred in parallel to the ASP, the relative importance of them was estimated.
Material and Methods
Study design and setting
Observational study that compared pre (20052008) and post ASP (2009 and 2015) periods. It was performed at the Hospital Militar of Santiago, a centre that attends active and retired military personnal and their relatives, as well as general public.
ASP
The ASP begun at year 2009 and was performed by 3 ID specialists during working days using a list of restricted antibiotics prepared daily by the Pharmacy Unit. The program was executed as a direct clinical visit using a prospective audit format with feedback to physicians in charge. The list of restricted molecules include 2nd to 4th generation cephalosporins, parenteral beta-lactam/beta-lactam inhibitors, parenteral quinolones, aminoglycosides, clindamycin, vancomycin, linezolid, tigecycline, antifungals and parenteral antiviral compounds. Supervision by ID specialists was allowed to stop or modify a prescription. The ASP included 3 courses on the period on antimicrobial use directed to medical staff and weekly seminars organized by the ID Unit. Besides, resident physicians in internal medicine were required to rotate during two months at the ID Unit to enhance their training. Most of them made night shifts taking initial decisions on therapy and were also in charge of medical wards. Only one guideline was available during the study period (surgical antimicrobial prophylaxis).
Antimicrobial consumption
Consumption was evaluated only in adult patients by daily defined doses (DDD) per 100 occupied beds, and sepatared by general wards and critical and intermediate care beds (CCU beds). Psychiatrics beds were excluded. Specific DDD were obtained at the WHO Collaborating Centre for Drug Statistics Methodology12. Bed occupancy was provided by the Biostatistical Unit of the hospital. Comparison was made by median values from the pre ASP era (2005 to 2008) with those of years 2009 and 2015 (ASP period). Data from intermediate years was not available (2011 to 2014). Antituberculosis and antiretrovirals drugs were not included because they are dispensed by a central governamental office. Trends were analyzed by Spearman correlation coefficient.
Expenditure
Data on antimicrobial charges, including taxes, was available from 2006. Values were upgraded to june 2015 using an online software from the Instituto Nacional de Estadísticas of Chile and converted to US dollars taken the parity change of the same month (1 US$ = 629.99 CLP). We calculated the annual expenditure per occupied bed (quotient of expenditure per bed occupancy) and compared variation as percentual changes from the basal value (median value from the pre ASP period). Trends were analyzed by Spearman correlation coefficient.
Antimicrobial resistance and nosocomial rates
To avoid contamination of data with community-acquired infections, we only included strains detected by active surveillance of nosocomial infections with only one isolate per patient, and excluding carrier cases. This information was available from 2010. Percentual changes were explored for methicillin-resistant Staphylococcus aureus, third generation-resistant Escherichia coli and Klebsiella pneumoniae isolates, as well as, carbapenem-resistant Pseudomonas aeruginosa strains. Changes in density rate (number of resistant isolates per occupied beds) were also analyzed. In all cases, trends were analyzed by Spearman correlation coefficient. Changes in nosocomial rates were also analyzed as detected by active surveillance of central venous catheter-associated bloodstream infections, ventilator-associated pneumonia, bloodstream infection related to hemodialysis catheter, urinary tract infection associated to indwelling catheter, vaginal or cesarean postpartum endometritis, and wound infections for clean or clean-contaminated surgical procedures at different sites. Besides, density rates of new events of Clostridium difficile diarrhea (first episode) and clinical infections associated to vancomycin-resistant Enterococci (VRE) were registered. Trends in nosocomial infections were analyzed by Spearman correlation coefficient.
Competitive biddings
Four open calls were issued between 2008 and 2014 for competitive biddings in order to provide all the pharmaceutical products used at the hospital. Selection was performed by predefined criteria looking for the best quality/price relationship. Only prequalified providers were allowed to participate.
Global mortality and mortality associated to infectious diseases codes
A possible effect of the ASP on global mortality or that associated to ID according to the International Classification of Diseases 10th revision (ICD codes A00-B99), was analyzed by adult death rates by 1000 discharges and trends analyzed. Information on deceased patients and specific ICD was obtained from the Biostatistics Unit. Discharges from the psychiatric unit were excluded. Data was available from 2005 to 2015.
Results
Antimicrobial consumption
Median basal consumption decreased from 221.3 DDD/100 beds to 170 DDD/100 beds after the ASP begun (Table 1). Consumption maintained the decreasing trend after the first year (-18.8%), and towards 2015, global consumption had declined 27.6% when compared with the median basal value. Reduction was observed both at CCU and general wards, but were not significant (Table 1). Falling was driven by decreasing trends in antibiotic consumption and antifungal consumption (Tables 1 and 2). Antivirals consumption did not change appreciably (median consumption 0.62 to 0.58 DDD/100 beds).
Compound | Median pre ASP* | Median post ASP | Median difference** | Spearman correlation coefficient r | p |
---|---|---|---|---|---|
All | 221.3 | 170.0 | -51.3 | -0.48 | ns*** |
General wards | 65.3 | 55.0 | -10.3 | -0.77 | ns |
CCU beds | 157.2 | 115.0 | -42.2 | -0.48 | ns |
All antibiotics | 211.8 | 139.2 | -72.6 | -0.48 | ns |
Antibiotics at CCU | 147.7 | 110.2 | -37.5 | -0.42 | ns |
Antibiotics at wards | 63.0 | 53.3 | -9.7 | -0.77 | ns |
All beta-lactams | 114.4 | 62.8 | -51.6 | -0.82 | < 0.05 |
Penicillins | 29.1 | 24.6 | -4.5 | -0.65 | ns |
Piperacillin-tazobactam | 4.6 | 10.2 | + 5,6 | + 0.94 | < 0.05 |
All Cephalosporins | 68.5 | 50.1 | -18.4 | -1.00 | ns |
1st G Cephalosporins | 20.5 | 13.1 | -7.4 | -1.00 | ns |
2nd G Cephalosporins | 2.4 | 0.5 | -1.9 | -0.94 | < 0.05 |
3rd G Cephalosporins | 35.8 | 35.9 | + 0.1 | +0.54 | ns |
4th G Cephalosporins | 9.8 | 0.7 | -9.1 | -1.00 | ns |
All carbapenems | 16.3 | 14.1 | -2.2 | +0.08 | ns |
Imipenem | 10.6 | 4.2 | -6.4 | -0.48 | ns |
Meropenem | 3.6 | 8.5 | +4.9 | +0.46 | ns |
Ertapenem | 2.4 | 1.4 | -1.0 | -0.42 | ns |
All quinolones | 27.6 | 18.3 | -9.3 | -0.48 | ns |
All aminoglycosides | 3.8 | 5.6 | +1.8 | +0.14 | ns |
All glycopeptides | 15.3 | 10.4 | -4.9 | -0.48 | ns |
All macrolides | 2.5 | 0.9 | -1.6 | -0.82 | < 0.05 |
Clindamycin | 8.8 | 6.2 | -2.6 | -0.88 | < 0.05 |
Metronidazole | 14.4 | 13.8 | -0.6 | -0.48 | ns |
Cotrimoxazole | 15.0 | 11.4 | -3.6 | +0.25 | ns |
Linezolid | 0.4 | 1.0 | +0.6 | +0.72 | ns |
Colistin | 0.7 | 2.0 | + 1.3 | + 0.65 | ns |
Tigecycline | 0.1 | 1.2 | + 1.1 | + 0.92 | < 0.05 |
*ASP: Antimicrobial stewardship program;
**a negative value indicates a reduction in consumption; a positive value an increase:
***ns: non significant.
Compound | Median Pre ASP* | Median Post ASP | Median difference** | Spearman | p |
---|---|---|---|---|---|
All antifungal | 9.0 | 6.2 | -2.8 | -0.65 | ns*** |
Antifungal at CCU | 7.0 | 4.6 | -2.4 | -0.20 | ns |
Antifungal at wards | 1.7 | 1.6 | -0.1 | -0.54 | ns |
All Echinocandines | 2.4 | 0.9 | -1.5 | -0.82 | < 0.05 |
Caspofungin | 2.4 | 0.0 | -2.4 | -0.98 | < 0.05 |
Anidulafungin | 0.0 | 0.9 | + 0.9 | +0.94 | < 0.05 |
Amphotericin deoxycolate | 0.6 | 0.3 | -0.3 | -0.82 | < 0.05 |
Fluconazole | 6.3 | 5.0 | -1.3 | -0.17 | ns |
Voriconazole | 0.1 | 0.0 | -0.1 | -0.65 | ns |
*ASP: Antimicrobial stewardship program;
**a negative value indicates a reduction in consumption; a positive value an increase;
***ns: non significant.
Analysis on antibiotic showed significant reductions among beta-lactams (median difference −51.6 DDD/100 beds, r = −0.87) and macrolides (median difference −1.6 DDD/100 beds, r = −0.82; Table 1 and Figure 1). Among beta-lactams, a marked decreasing trend was observed for first and fourth generation cephalosporins, ertapenem, imipenem, penicillins, quinolones, glycopeptides, metronidazole, and cotrimoxazole (Table 1, Figure 1). Clindamycin and second generation cephalosporins consumption decreased significantly.
During the study period, consumption significantly increased for piperacillin-tazobactam (median difference +5.6 DdD/100 beds, r = +0.94), and tygecyline (median difference +1.1 DDD/100 beds, r = +0.93). This was accompanied by similar but not significant trends for third generation cephalosporins, meropenem, aminoglycosides, linezolid, and colistin (Table 1, Figure 1). A replacement phenomenon on carbapenems preference was perceived: increases in meropenem consumption were accompanied by a parallel reduction in ertapenem and imipenem prescription with no global expansion of these compounds (Table 1, Figure 1).
Antifungals prescription preferences also changed. Caspofungin consumption was totally replaced by anidulafungin (Table 2, Figure 2). Voriconazole and fluconazole consumption showed decreasing trends. Amphotericin deoxycolate consumption decreased.
Nosocomial infection rates
Most nosocomial infection rates under surveillance in adult patients did not demonstrate significant variations during the analyzed period. They included those associated to invasive procedures, vaginal or cesarean postpartum endometritis, and wound infections for clean or clean-contaminated surgical procedures at different sites (data not shown). Only infections associated to total hip replacement significantly rise between 2005 and 2015 (2.1% to 3.3%; r = +0.94, p < 0.05). These results suggest that variation in nosocomial infection rates appear not have induced antimicrobial consumption modifications. On the other hand, density rate of new events of diarrhea associated to C. difficile (0.4 to 0.1 events per 1000 bed-days between 2010 and 2015; r = −0.88, p < 0.05) and clinical infections linked to VRE (0.14 to 0.04 between 2003 and 2015; r = −0.71), significantly lowered during or after ASP begun.
Economic impact
Global antimicrobial expenditure per bed plummeted from a median of 13 to 10 US$ at the second year (2010, −28%), and to US$ 6 during year 2015 (-57%). For CCU, values descended from a median of US$ 59 to 36 per bed at 2010 (-23%), and to US$ 15 per bed at year 2015 (-75%). For general wards, changes were similar: 7 to 6 US$ per bed (-12%) at 2010 and to US$ 5 per bed at 2015 (-32%). These variations were significant for CCU (r = −0.90, p < 0.05). However, not all this saving is attributable to reduced antimicrobial consumption. Consumption decreased 27.6% but expenditure lower further (-57%) at the last year of the study. We assumed that only a fraction of cost-saving were related to the impact on consumption (the difference by biddings), and calculated this factor at the last year as the proportion obtained from the quotient consumption/ cost-saving (-27.6%/-57%), given an estimate of 48.4% as a direct contribution of the ASP. That means that at the last year of follow up, approximately half of the impact on expenditure was explained by the control program, and the other half by competitive biddings (Figure 3).
The next step was to estimate cost-saving in currency. Antimicrobial charges per occupied bed from 2005 to 2008 were taken as the basal value (median US$ 13 per bed). This cost was multiplied by the number of total occupied beds, allowing estimate the pre ASP global charge on antimicrobials. The same formula was applied for 2015 using this time US$ 6 per occupied bed. Using this approach, it was estimated that during the last year, cost-saving reached 393 072 US$. Nonetheless, only 48.4% of this amount was estimatively attributable to the ASP, an amount equivalent to 190 000 US$.
Antimicrobial resistance trends
During years 2010 to 2015, 42 S. aureus, 40 E. coli, 23 K. pneumoniae, and 41 P. aeruginosa isolates were identified associated to different nosocomial infections. Analysis of percentual changes of resistant isolates did not show significant variations (data not shown). Density rates (number of resistant isolates belonging to any one of these groups expressed by occupied beds), also remain stationary (data not shown). No information was available from the pre ASP era.
Global mortality and infectious diseases-associated mortality trends
From 2005 to 2015, a significant reducing trend was observed both for global mortality (31 to 25 per 1000 discharges; r= −0.72, p < 0.05) and infectious diseases-associated mortality (4 to 2 per 1000 discharges; r= −0.81, p < 0.05, Figure 4).
Discussion
Our results demonstrate that our ASP was associated with a favorable impact on antimicrobial consumption, cost-saving and ID-related mortality. The effect intensified along time and involved different compound families. Results on expenditure were more pronounced than those attributable to consumption reduction, indicating an additive effect of sequential competitive biddings. Approximately, for every two dollars saved, one was achieved by the ASP. Besides, as salaries of the 3 ID specialists involved in control reached US$ 130 000 per year and savings were of a higher magnitude the program itself was sustainable. This information is useful to convince hospital administrators in order to support initial hiring of dedicated human resources dedicated to ASP as has been shown in other reports5,7,8,13–22. To the best of our knowledge, this is the first report that analyzes the contributory role of bidding and ASP at the same time. In addition, we believe our results are relevant because information on the long-term effect of ASP is scarce9,23, and as shown here, the impact appears to deepen along time. In our hospital, consumption decline appears not to be linked with a parallel reduction in most nosocomial infection rates.
Consumption increase involved few molecules, and probably by desired reasons. Piperacillin-tazobactam was frequently used in CCU patients in prolonged infusion24, replacing alternatives. Tygecycline use in polymicrobial multiresistant infections, avoided simultaneous use of carbapenems and glycopeptides (off label indication)25. On the other hand reduction in caspofungin and ertapenem utilization obeyed to their higher cost in comparison with suitable alternatives in their respective families.
Despite data on antimicrobial resistance was not available from the pre ASP era, trend analysis did not evidence a declining pattern for none of the species. In contrast, most of the literature describes antimicrobial resistance among inpatient-associated isolates without indicating how much of them could be related to readmitted or transferred patients. The declining curve on C. difficile- associated diarrhea or clinical infections linked to VRE could be interpreted as a successful accomplishment of the ASP. Nonetheless, both problems were specifically managed by the nosocomial infection control group with dedicated intervention programs that started almost in parallel to the ASP, turning difficult to read the contribution of each separately.
A favorable impact of ASP on mortality has rarely been reported for ICU or medical ward patients9,26, and our results suggests an additional benefit of these programs. The effect was observed on global mortality and explained by a significant impact on infectious disease-related deaths. Several reports and reviews support a beneficial impact of ID consultations on mortality through several scenarios in general hospital or ICU settings, S. aureus or Candida bloodstream infections, infectious endocarditis or infectious complications on solid organ transplant patients27–32.
This work has some limitations. Data on consumption on intermediate years after the ASP begun was not available, making theoretically possible years with higher antimicrobial prescription. We believe this is improbable because the working format remain, the effect augmented along time and expenditure also plummeted. The hospital moved to a new building on 2009, the same year ASP begun. The new facility diminished patient numbers per room (from 6 to 3 or 2), a fact that could have decrease nosocomial infection rates and antimicrobial use. Our results suggest that this explanation is highly improbable because neither nosocomial infections rates nor antimicrobial resistant isolates change appreciably during the study period. Aditionally, a transversal impact over several families of antimicrobial compounds was observed, not only for those applied in nosocomial infections management, and the positive impact was maintained and deepened several years after the hospital moved out. In addition, it is unknown if more educational activities, more antimicrobial guidelines or better diagnostics technologies could have had a more pronounced impact on consumption. Furthermore, severity scores were not included in our work and it is possibly that less severely ill patients predominate after ASP begun. If that were the case, non-infectious related mortality should have also fade away, but this phenomenon did not occur making unlikely this explanation.
In conclusion, our ASP was associated in the short- and long-term with a favorable impact on antimicrobial consumption, expenditure, and decreased mortality. Parallel competitive biddings amplified control expenditure obtained by the ASP. The program was sustainable as salaries of ID specialists participating in the ASP were lower than savings obtained. Using only information linked to nosocomial isolates, we could not demonstrate a reversion in antimicrobial resistance.