Defining Appropriate Use of Proton-Pump Inhibitors Among Medical Inpatients
J Gen Intern Med
Defining Appropriate Use of Proton-Pump Inhibitors Among Medical Inpatients
Matt Pappas 0 1 2
0 The University of Michigan , Ann Arbor, MI , USA
1 Department of Internal Medicine, Division of General Internal Medicine, The University of Michigan Health System , Ann Arbor, MI , USA
2 VA Center for Clinical Management Research, VA Ann Arbor Healthcare System , Ann Arbor, MI , USA
BACKGROUND: Proton-pump inhibitors (PPIs) are commonly used among medical inpatients, both for prophylaxis against upper gastrointestinal bleeding (UGIB) and continuation of outpatient use. While PPIs reduce the risk of UGIB, they also appear to increase the risk of hospitalacquired pneumonia (HAP) and Clostridium difficile infection (CDI). Depending upon the underlying risks of these conditions and the changes in those risks with PPIs, use of proton-pump inhibitors may lead to a net benefit or net harm among medical inpatients. OBJECTIVE: We aimed to determine the net impact of PPIs on hospital mortality among medical inpatients. DESIGN: A microsimulation model, using literaturederived estimates of the risks of UGIB, HAP, and CDI among medical inpatients, along with the changes in risk associated with PPI use for each of these outcomes. The primary outcome was change in inpatient mortality. PARTICIPANTS: Simulated general medical inpatients outside the intensive care unit (ICU). MAIN MEASURE: Change in overall mortality during hospitalization. KEY RESULTS: New initiation of PPI therapy led to an increase in hospital mortality in about 90 % of simulated patients. Continuation of outpatient PPI therapy on admission led to net increase in hospital mortality in 79 % of simulated patients. Results were robust to both one-way and multivariate sensitivity analyses, with net harm occurring in at least two-thirds of patients in all scenarios. CONCLUSIONS: For the majority of medical inpatients outside the ICU, use of PPIs likely leads to a net increase in hospital mortality. Even in patients at particularly high risk of UGIB, only those at the very lowest risk of HCAP and CDI should be considered for prophylactic PPI use. Continuation of outpatient PPIs may also increase expected hospital mortality. Apart from patients with active UGIB, use of PPIs in hospitalized patients should be discouraged.
simulation; modeling; hospital medicine; medical decision making
Proton-pump inhibitors (PPIs) are commonly used among
medical inpatients. An estimated 50 % of inpatients are prescribed
PPIs, including between 15 and 25 % who are prescribed these
agents specifically for prophylaxis of upper gastrointestinal
bleeding (UGIB).1–4 However, only two small controlled trials
of prophylaxis exist in the inpatient population outside of the
intensive care unit (ICU).4 Thus, evidence to support this
practice is largely extrapolated from studies carried out within ICUs,
where risk of UGIB is substantially greater due to the frequent
presence of respiratory failure.5 Nevertheless, it is reasonable to
conclude that PPIs do reduce the risk of clinically significant
UGIB among non-ICU inpatients.
Unfortunately, enthusiasm for routine use if PPIs is
tempered by evidence that PPIs are associated with increased risk
of hospital-acquired pneumonia (HAP) and Clostridium
difficile infection (CDI).1,6,7 Because these conditions are
common and often have worse outcomes than UGIB, many
authors have suggested that prophylactic PPI use should be
avoided in most hospitalized patients.2,3,7,8 Overuse is
sufficiently common that, as one of its recommendations for the
American Board of Internal Medicine (ABIM) Foundation’s
Choosing Wisely campaign, the Society of Hospital Medicine
recommended against stress ulcer prophylaxis “unless at high
risk for GI complications”.9 However, there are no published
analyses examining how the tradeoffs between UGIB risk and
HAP and CDI risk affect the net effect of acid suppression on
inpatient mortality, nor are there studies examining how
variation in patient risk of UGIB, HAP, and CDI should affect
individual decisions about use of PPIs. The different
underlying risks of UGIB, HAP, and CDI, along with the different
relative risks for each of these conditions with PPI use, may
lead to subpopulations of inpatients in whom PPI use may
either increase or decrease overall mortality. Using modeling
and simulation techniques and literature-derived distributions
of the risk and case fatality of the major conditions affected by
PPI use, we sought to examine the overall impact of the two
most common indications for inpatient PPI use—inpatient
stress ulcer prophylaxis and continuation of outpatient
use—on inpatient mortality. In addition, we sought to
better define populations of inpatients outside the ICU
for whom each type of use might yield a net benefit or
We created a microsimulation model to examine the effect of
PPI continuation or initiation on in-hospital mortality among
medical inpatients outside the ICU. To estimate the impact of
PPI use, we modeled the risk of each of the three outcomes of
interest (UGIB, HAP, and CDI), and in our base case analysis,
assumed a causal linkage between PPI use and the risk of each
of these outcomes.
A causal link between PPI use and reduction in UGIB risk is
well-supported both in theory and published data from ICU
studies,5,10–12 although there are limited randomized
controlled trial data outside that setting.4 The causal linkages
between acid suppressive therapy and each of HAP and CDI
are perhaps less well established. For each of these conditions,
observational studies have shown clear and consistent
associations with acid suppressive therapy, and there appear to be
dose-response relationships and clear temporal relationships
between initiation of therapy and risk of both HAP and
CDI.6,7,13–18 Additionally, separate work in healthy volunteers
has delineated a plausible biologic mechanism by which acid
suppression would increase the risk of pneumonia.19,20 To our
knowledge, investigations into the mechanism through which
PPIs increase risk of CDI have been limited to animal models
and culture data, and some conflict remains.21–23 Thus, while
not definitively proven in clinical trials, a preponderance of
observational evidence appears to support a causal linkage
between acid suppression and risk of both HAP and CDI, with
the mechanistic linkage between PPIs and CDI somewhat less
SOURCES OF MODEL PROBABILITIES
We first sought to identify the best available literature
estimates and distribution of the incidence of UGIB, HAP, and
CDI, the case fatality associated with UGIB, HAP, and CDI,
and the odds ratios conferred upon UGIB, HAP, and CDI by
initiation or continuation of PPI therapy. To do so, we
conducted a literature search using the MEDLINE database, and
used data from the Healthcare Cost and Utilization Project’s
National Inpatient Sample (HCUP NIS) to verify and
supplement rates found in the literature;24 further detail regarding our
literature search strategy can be found in the Online appendix.
The risks of UGIB, HAP, and CDI varied widely in the
identified literature. Estimates of nosocomial UGIB risk in
non-ICU patients range from 0.22 % to 0.4 %, depending
upon the cohort and definition of UGIB used (e.g.,
endoscopically proven vs. hematemesis or melena vs. occult bleeding
with drop in hemoglobin).1,25 Estimates of the incidence of
HAP outside the ICU range between 0.3 % and 4.9 %, again
varying by cohort and case definition.6,26,27 Point estimates of
the risk of hospital-acquired CDI in academic medical centers
range from 0.7 % to 2.25 % (during an outbreak).7,28–31
Surveillance data from Canada (where CDI is a reportable
infection) suggest an overall rate of 0.46 %, with 95 %
confidence intervals from 0.34 % to 0.84 %.32 Some portion of this
risk is attributable to use of acid suppressive medications,
though CDI incidence among patients not taking either PPIs
or histamine2-receptor antagonists is infrequently reported.
Howell and colleagues report the CDI incidence at their center
among patients not prescribed any form of acid suppressive
therapy to be 0.3 %.7 Many authors reported incidence rates
rather than overall incidence; these ranged from 6.5 to 28.1
cases per 10,000 patient-days, with an overall trend that
appears to be increasing over time.29,33,34
Outcomes of UGIB, HAP, and CDI among inpatients were
similarly varied in our literature search. For example, in data
validating the Rockall score for risk prediction in UGIB,
mortality ranged from 0 % to 46.5 %, depending on various
risk characteristics.35,36 HAP was associated with in-hospital
mortality of 18.4 % in an Italian cohort, 18.8 % mortality in a
United States cohort, and 26 % (with 14 % mortality attributed
to pneumonia) in a study of patients in Spain.15,26,37
Hospitalacquired CDI was associated with an 11 % absolute increase in
hospital mortality in one study; another found that attributable
mortality varied with age from 0 to 14.0 %.38,39 Dubberke and
colleagues reported attributable mortality after 180 days to be
The impact of acid suppression therapy on each of the
considered conditions varied as well. Herzig and colleagues
found adjusted odds ratios with PPI use of 0.58 and 1.3 for
nosocomial GI bleed and HAP, respectively.1,6 Using the same
data set, Howell and colleagues found odds ratios for CDI
ranging from 1.53 to 2.36, with more intensive acid
suppression associated with higher risk of CDI.7 Meanwhile, two
meta-analyses found pooled odds ratios of CDI with PPI use
of 1.74 and 1.69.41,42
Studies of outpatient PPI use have demonstrated a possible
diminishing risk of pneumonia with longer durations of acid
suppressive therapy.14,43 Because similar data are not available
for discontinuation of PPIs when chronic users are
hospitalized, we assumed that the lowest odds ratio observed among
outpatients would apply to discontinuation among inpatients.
That is, we assumed that chronic users who continued to
receive PPIs during hospitalization would be subject to an
odds ratio of HAP of 1.09, compared to a mean OR of 1.28
with new PPI initiation.14
For each of our parameters, we used literature estimates that
included information regarding variance (quartiles, deciles,
confidence intervals, etc.), rather than point estimates,
ensuring that our ranges for each parameter included other estimates
identified in our literature search. The base case and
distributions for each major assumption are shown in Table 1. We
modeled most parameters as normal distributions, with mean
and standard deviation to yield a distribution matching that of
reported confidence intervals. This was the case for the
incidence of HAP, case fatality rates of HAP and CDI, and odds
ratios of UGIB, HAP, and CDI.6,7,15 If more detailed
information on distributions was available, we used the distribution
that best fit the published estimates. For example, a Weibull
distribution was created to match Herzig and colleagues’
reported incidence of UGIB.44 To estimate the case fatality
rate of UGIB, a normal distribution was fitted to Vreeburg and
colleagues’ reported mean and standard deviation Rockall
score.36 This distribution was used to generate Rockall scores
as a categorical variable, yielding an expected frequency of
UGIBs of each severity. As described above, many definitions
of UGIB are used in the literature; by adopting the data
published by Rockall and colleagues, we also adopted their
definition. The combined Rockall and Vreeburg validation set
was used to estimate mortality within each subgroup.36
Estimates of CDI incidence are complicated by increasing rates
over time and wide variation among hospitals and regions.
Because we were unable to identify recent, nationally
representative estimates of CDI incidence with information
regarding variance, we estimated this distribution from the NIS
database. We used multiple methods, finding convergence
between methods of estimation. Further detail regarding this
incidence estimate can be found in the Online appendix. The
mean, median, 5th, and 95th percentile of probability
distributions for key variables from a representative simulation are
included in Table 1.
We then performed a Monte Carlo simulation, estimating the
change in hospital mortality from PPI initiation or continuation
for hypothetical cohorts of patients sampled from the
abovedetailed probability distributions. Given the somewhat less
robust mechanistic evidence linking acid suppressive therapy
and CDI, we performed separate simulations with and without
an effect of PPIs on CDI risk. Because our focus was on PPI use
among inpatients, we limited our analysis to hospital mortality,
and did not consider potential effects on other outcomes
associated with longer-term use of PPI therapy, such as B-12
deficiency, hip fracture, or altered medication absorption.8
Version 13 of Stata was used to analyze the NIS dataset and
obtain a distribution of CDI. All other analysis and simulation
was performed in R (version 3.1).
The primary results of our simulation are shown in Figures 1
and 2. Assuming a causal link between PPI use and CDI, in
approximately 90 % of simulated cases, initiation of PPIs for
prophylaxis led to a net increase in expected mortality, and
therefore likely represents harm to the overwhelming majority
of hospitalized patients. When we simulated the alternate case
where PPIs do not contribute to risk of CDI, new initiation of
PPI therapy leads to expected net harm in approximately 86 %
of hospitalized patients. A density plot, showing the
distribution of impact on expected mortality under these two
assumptions, is shown in Fig. 1. In our base-case analysis, the
number-needed-to-harm (the number of patients who would
be newly initiated on a PPI for prophylaxis to cause one
additional inpatient death) is approximately 830.
Our simulation of PPI discontinuation among chronic users
during hospitalization showed that continuation of PPIs would
lead to increase in expected mortality in approximately 80 %
of patients, and in 68 % of patients in the case where PPIs do
not contribute to risk of CDI. A density plot showing the
results of our simulation under these assumptions is shown
in Fig. 2.
To better define the effects of prophylactic PPI use on
specific risk groups within the inpatient population, we
prepared tables examining the net effect on mortality of PPI use
across the probabilities of each complication. The base case
estimate (allowing the probability of HAP, case fatality of
UGIB, case fatality of HAP, and case fatality of CDI all to
be at the 50th percentile) is shown in Table 2A; at all deciles of
CDI and UGIB risk, prophylactic PPI use would lead to net
harm. A more favorable case for PPI prophylaxis emerges in
Table 2B, wherein we have used the 10th percentile for
probability of HAP, case fatality of HAP, and case fatality of CDI,
and the 90th percentile of UGIB case fatality. In this case,
which is strongly biased in favor of PPI use, approximately
half of patients would benefit from PPI prophylaxis. Put
another way, even in a scenario that is skewed to heavily favor
the prophylactic use of PPIs, the net effect on inpatient
mortality is zero.
In our base case scenario, only extreme outliers in
underlying risks stand to benefit from PPI prophylaxis. Assuming a
HAP risk at the median, even a patient in the highest decile of
Fig. 1 Density plot of change in expected mortality with newly-initiated PPI, assuming increased risk of CDI with acid suppression (black) and
assuming no increased risk of CDI (gray). The mean impact of PPI initiation on expected mortality is indicated by dashed vertical lines. Patients
to the left of zero receive a net benefit, while patients to the right of zero are harmed.
UGIB risk and lowest decile of CDI risk stands to be harmed
(Table 2). At the lowest decile of HAP risk, only patients in the
highest decile of UGIB risk stand to receive benefit (not
shown). In contrast to these examples, the typical inpatient is
likely to be harmed by prophylactic use of PPIs.
Our study suggests that for the vast majority of medical
inpatients, routine use of PPI therapy for UGIB prophylaxis
is harmful, leading to a net increase in expected mortality.
These results were robust to a wide variety of parameters and
risk distributions. This strongly suggests that prophylaxis with
PPIs is unwarranted in most patients. Additionally, we found
that continuation of outpatient PPI use during hospitalization
may also contribute to increased mortality. This finding
suggests that discontinuation or down-titration of PPIs during
hospitalization should be considered in patients who
chronically take PPIs but are not at extremely high risk of UGIB
(e.g., the majority of patients who take PPIs for symptoms of
reflux or as prophylaxis while taking NSAIDs, steroids, or
other medications of concern).
Our findings raise concern, because overuse of PPIs is
common among inpatients, and this highlights the need for
active methods of ensuring appropriate use. Even after
interventions designed to be specifically intended to reduce PPI
prescriptions among hospitalized patients, PPI use appears to
be common: after such an intervention, Yachimski and
colleagues reported that 16 % of admitted patients not taking a
PPI on admission were prescribed PPIs during their
hospitalization.3 In our model, approximately 10 % of patients derive
benefit from prophylactic PPI therapy, suggesting that
inappropriate use of these medications as prophylaxis outstrips
helpful use by at least 60 %, even under the most generous
of assumptions and following interventions designed to reduce
While PPIs are clearly overused, our analysis also
suggests that a small, very high-risk segment of
inpatients may be well-served by prophylactic acid
suppression. As with the outliers discussed above and shown in
Table 2, patients stand to benefit from prophylactic acid
suppression if the risk of UGIB and mortality is high
and the risk of CDI and HAP (and associated mortality)
are low. Identifying such patients requires careful risk
prediction, and commonly used heuristics (such as using
PPIs in those who are prescribed steroids or NSAIDs)
do not capture the tradeoffs that must be evaluated to
optimally target acid suppressive therapy. Our risk tables
suggest that increased risk of UGIB alone is not
adequate; instead, identifying patients who may benefit
requires prospective risk prediction tools for risks and
fatality rates of UGIB, HAP, and CDI. The existing risk
models we identified have not been fully validated.
Further, many published risk prediction tools require
data and calculations that would be best addressed by
automated calculation performed in electronic medical
records/order entry systems, rather than by individual
Finally, our analysis raises questions regarding
continuation of outpatient acid suppressive therapies. In our
simulation of this scenario, continuing use of acid
suppressive therapy among those hospitalized would lead to
a small increase in expected mortality in around 80 %
of patients. Even under the assumptions that PPIs do not
contribute to increased risk of CDI and that continuation
of outpatient therapy leads to a lesser increase in risk of
HAP compared with new initiation, continuation of
outpatient therapy appears to lead to a small increase
in expected mortality in over two-thirds of patients.
Unfortunately, withholding long-term PPIs during
hospitalization may lead to an increase in dyspeptic
symptoms.45 This particular action may therefore require
setting appropriate patient expectations, and tolerance of
said dyspeptic symptoms; alternatively, patients with
severe dyspepsia may be willing to accept small risks
of adverse events in order to minimize their symptoms.
Ultimately, as with most decisions involving tradeoffs
between symptomatic benefit and potential adverse
events, individualized decision-making is warranted.
Alternatively, for patients with severe rebound
symptoms, hospital admission may be an opportunity to
down-titrate acid suppressive regimens (e.g., to lower
doses, to potentially less-harmful histamine-2 antagonists, or
to as-needed antacids).
Our study is not without limitations. First, this is a
modeling study, and is therefore reliant on
literaturederived estimates of the risks and mortality associated
with the conditions considered. Second, our distributions
are fitted to published estimates, and may not be
All numbers in percents. Negative numbers (shaded cells) indicate a net benefit is conferred by starting a PPI.
Table A (top) shows the base case (that is, assuming the mean for all of: probability of HAP, case fatality of UGIB, case fatality of HAP, and case fatality
of CDI). Due to the skew in the distribution of UGIB mortality, using the mean rather than the median for UGIB case fatality overestimates the benefit of
PPIs compared with the median. Using the median here would more accurately predict change in mortality for an average patient, but because we are
estimating more favorable scenarios for PPI use, we have used the mean.
Table B (bottom) shows the 10th percentile for risk and case fatality of HAP, 10th percentile for case fatality of CDI, and 90th percentile for case fatality
of UGIB. This represents patients both high risk for UGIB of high mortality and equally low risk of HAP and mortality from HAP and CDI.
representative of the patient population at any particular
institution. Our results should be viewed in that context,
and interpreted through the lens of local risks of these
conditions. And finally, as discussed above, prospective
risk assessments for UGIB, HAP, and CDI are limited at
present. Available predictive models of UGIB have been
retrospectively validated in multiple cohorts, but have
not been prospectively validated. CDI risk prediction
models have not been prospectively validated, and often
rely on difficult-to-measure factors such as exposure to
Clostridium difficile. To our knowledge, no models
predictive of HAP risk have been validated.
Despite these limitations, we posit that the use of
PPIs either for inpatient prophylaxis or continuation of
outpatient use likely leads to a small increase in
expected hospital mortality for the majority of medical
inpatients. Targeted use of PPIs to the small proportion
of patients who may benefit is theoretically possible, but
requires the development and validation of accurate risk
prediction tools for risks and mortality from UGIB,
HAP, and CDI. Given the current lack of such tools, our
results suggest that, excepting those who are being
actively treated for UGIB, PPI use among general
medical inpatients should generally be discouraged, even in
those who have been taking them chronically.
Acknowledgements: The authors wish to thank Andrew Odden, MD,
for his helpful comments on an earlier version of this manuscript.
Corresponding Author: Matt Pappas, MD, MPH; The University of
Michigan, Ann Arbor, MI, USA (e-mail: ).
Compliance with Ethical Standards:
Conflicts of interest: The authors declare that they do not have a
conflict of interest.
1. Herzig SJ , Vaughn BP , Howell MD , Ngo LH , Marcantonio ER . Acidsuppressive medication use and the risk for nosocomial gastrointestinal tract bleeding . Arch Intern Med . 2011 ; 171 ( 11 ): 991 - 997 . doi: 10 .1001/ archinternmed. 2011 . 14 .
2. Nardino RJ , Vender RJ , Herbert PN . Overuse of acid-suppressive therapy in hospitalized patients . Am J Gastroenterol . 2000 ; 95 ( 11 ): 3118 - 3122 . doi: 10 .1111/j.1572- 0241 . 2000 . 03259 .x.
3. Yachimski PS , Farrell EA , Hunt DP , Reid AE . Proton pump inhibitors for prophylaxis of nosocomial upper gastrointestinal tract bleeding: effect of standardized guidelines on prescribing practice . Arch Intern Med . 2010 ; 170 ( 9 ): 779 - 783 . doi: 10 .1001/archinternmed. 2010 . 51 .
4. Janicki T , Stewart S . Stress-ulcer prophylaxis for general medical patients: a review of the evidence . J Hosp Med . 2007 ; 2 ( 2 ): 86 - 92 . doi: 10 . 1002/jhm.177.
5. Cook DJ , Fuller HD , Guyatt GH , et al. Risk factors for gastrointestinal bleeding in critically ill patients . Canadian Critical Care Trials Group. N Engl J Med . 1994 ; 330 ( 6 ): 377 - 381 . doi: 10 .1056/NEJM199402103300601.
6. Herzig SJ . Acid-suppressive medication use and the risk for hospitalacquired pneumonia . JAMA . 2009 ; 301 ( 20 ): 2120 - 2128 . doi: 10 .1001/jama. 2009 . 722 .
7. Howell MD , Novack V , Grgurich P , et al. Iatrogenic gastric acid suppression and the risk of nosocomial Clostridium difficile infection . Arch Intern Med . 2010 ; 170 ( 9 ): 784 - 790 . doi: 10 .1001/archinternmed. 2010 . 89 .
8. Heidelbaugh JJ , Goldberg KL , Inadomi JM . Overutilization of proton pump inhibitors: a review of cost-effectiveness and risk [corrected] . Am J Gastroenterol . 2009 ; 104 ( Suppl 2 ): S27 - S32 . doi: 10 .1038/ajg. 2009 . 49 .
9. Bulger J , Nickel W , Messler J , et al. Choosing wisely in adult hospital medicine: five opportunities for improved healthcare value . J Hosp Med . 2013 ; 8 ( 9 ): 486 - 492 . doi: 10 .1002/jhm. 2063 .
10. Cook D , Heyland D , Griffith L , Cook R , Marshall J , Pagliarello J . Risk factors for clinically important upper gastrointestinal bleeding in patients requiring mechanical ventilation . Canadian Critical Care Trials Group. Crit Care Med . 1999 ; 27 ( 12 ): 2812 - 2817 .
11. Lin P-C , Chang C-H, Hsu P-I, Tseng P-L , Huang Y -B. The efficacy and safety of proton pump inhibitors vs histamine-2 receptor antagonists for stress ulcer bleeding prophylaxis among critical care patients: a metaanalysis . Crit Care Med . 2010 ; 38 ( 4 ): 1197 - 1205 . doi: 10 .1097/CCM. 0b013e3181d69ccf .
12. Pimentel M , Roberts DE , Bernstein CN , Hoppensack M , Duerksen DR . Clinically significant gastrointestinal bleeding in critically ill patients in an era of prophylaxis . Am J Gastroenterol . 2000 ; 95 ( 10 ): 2801 - 2806 . doi: 10 . 1111/j.1572- 0241 . 2000 . 03189 .x.
13. Dial S , Delaney JAC , Barkun AN , Suissa S . Use of gastric acidsuppressive agents and the risk of community-acquired Clostridium difficile-associated disease . JAMA . 2005 ; 294 ( 23 ): 2989 - 2995 . doi: 10 . 1001/jama.294.23.2989.
14. Sarkar M , Hennessy S , Yang Y -X. Proton-pump inhibitor use and the risk for community-acquired pneumonia . Ann Intern Med . 2008 ; 149 ( 6 ): 391 - 398 .
15. Venditti M , Falcone M , Corrao S , Licata G , Serra P , Study Group of the Italian Society of Internal Medicine. Outcomes of patients hospitalized with community-acquired, health care-associated, and hospital-acquired pneumonia . Ann Intern Med . 2009 ; 150 ( 1 ): 19 - 26 .
16. Eurich DT , Sadowski CA , Simpson SH , Marrie TJ , Majumdar SR . Recurrent community-acquired pneumonia in patients starting acidsuppressing drugs . Am J Med . 2010 ; 123 ( 1 ): 47 - 53 . doi: 10 .1016/j. amjmed. 2009 . 05 .032.
17. Dial S , Alrasadi K , Manoukian C , Huang A , Menzies D . Risk of Clostridium difficile diarrhea among hospital inpatients prescribed proton pump inhibitors: cohort and case-control studies . CMAJ . 2004 ; 171 ( 1 ): 33 - 38 . doi: 10 .1503/cmaj.1040876.
18. Linsky A , Gupta K , Lawler EV , Fonda JR , Hermos JA . Proton pump inhibitors and risk for recurrent Clostridium difficile infection PPIs and recurrent C difficile infection . Arch Intern Med . 2010 ; 170 ( 9 ): 772 - 778 . doi: 10 .1001/archinternmed. 2010 . 73 .
19. Thorens J , Froehlich F , Schwizer W , et al. Bacterial overgrowth during treatment with omeprazole compared with cimetidine: a prospective randomised double blind study . Gut . 1996 ; 39 ( 1 ): 54 - 59 .
20. du Moulin GC , Paterson DG , Hedley-Whyte J , Lisbon A . Aspiration of gastric bacteria in antacid-treated patients: a frequent cause of postoperative colonisation of the airway . Lancet . 1982 ; 1 ( 8266 ): 242 - 245 .
21. Jump RLP , Pultz MJ , Donskey CJ . Vegetative Clostridium difficile survives in room air on moist surfaces and in gastric contents with reduced acidity: a potential mechanism to explain the association between proton pump inhibitors and C. difficile-associated diarrhea? Antimicrob Agents Chemother . 2007 ; 51 ( 8 ): 2883 - 2887 . doi: 10 .1128/AAC.01443- 06 .
22. Nerandzic MM , Pultz MJ , Donskey CJ . Examination of potential mechanisms to explain the association between proton pump inhibitors and Clostridium difficile infection . Antimicrob Agents Chemother . 2009 ; 53 ( 10 ): 4133 - 4137 . doi: 10 .1128/AAC.00252- 09 .
23. Kaur S , Vaishnavi C , Prasad KK , Ray P , Kochhar R . Comparative role of antibiotic and proton pump inhibitor in experimental Clostridium difficile infection in mice . Microbiol Immunol . 2007 ; 51 ( 12 ): 1209 - 1214 .
24. Agency for Healthcare Research and Quality. HCUP National Inpatient Sample (NIS) . http://www.hcup-us.ahrq.gov/databases.jsp.
25. Qadeer MA , Richter JE , Brotman DJ . Hospital-acquired gastrointestinal bleeding outside the critical care unit: risk factors, role of acid suppression, and endoscopy findings . J Hosp Med . 2006 ; 1 ( 1 ): 13 - 20 . doi: 10 .1002/jhm.10.
26. Sopena N , Sabrià M , Neunos 2000 Study Group . Multicenter study of hos pi tal -a cq u ire d p neu moni a in no n-I CU p ati en ts . Ches t. 2005 ; 127 ( 1 ): 213 - 219 . doi: 10 .1378/chest.127.1.213.
27. Harkness GA , Bentley DW , Roghmann KJ . Risk factors for nosocomial pneumonia in the elderly . Am J Med . 1990 ; 89 ( 4 ): 457 - 463 .
28. Reveles KR , Lee GC , Boyd NK , Frei CR . The rise in Clostridium difficile infection incidence among hospitalized adults in the United States: 2001 - 2010 . Am J Infect Control . 2014 ; 42 ( 10 ): 1028 - 1032 . doi: 10 .1016/j.ajic. 2014 . 06 .011.
29. Dubberke ER , Yan Y , Reske KA , et al. Development and validation of a Clostridium difficile infection risk prediction model . Infect Control Hosp Epidemiol . 2011 ; 32 ( 4 ): 360 - 366 . doi: 10 .1086/658944.
30. Dubberke ER , Reske KA , Olsen MA , et al. Evaluation of Clostridium difficile-associated disease pressure as a risk factor for C difficile-associated disease . Arch Intern Med . 2007 ; 167 ( 10 ): 1092 - 1097 . doi: 10 .1001/archinte. 167.10.1092.
31. Loo VG , Poirier L , Miller MA , et al. A predominantly clonal multiinstitutional outbreak of Clostridium difficile-associated diarrhea with high morbidity and mortality . N Engl J Med . 2005 ; 353 ( 23 ): 2442 - 2449 . doi: 10 . 1056/NEJMoa051639.
32. Gravel D , Miller M , Simor A , et al. Health care-associated Clostridium difficile infection in adults admitted to acute care hospitals in Canada: a Canadian Nosocomial Infection Surveillance Program Study . Clin Infect Dis . 2009 ; 48 ( 5 ): 568 - 576 . doi: 10 .1086/596703.
33. Loo VG , Bourgault A-M , Poirier L , et al. Host and pathogen factors for Clostridium difficile infection and colonization . N Engl J Med . 2011 ; 365 ( 18 ): 1693 - 1703 . doi: 10 .1056/NEJMoa1012413.
34. Zilberberg MD , Shorr AF , Kollef MH . Increase in adult Clostridium difficile-related hospitalizations and case-fatality rate , United States , 2000 - 2005 . Emerging Infect Dis. 2008 ; 14 ( 6 ): 929 - 931 . doi: 10 .3201/ eid1406. 071447 .
35. Rockall TA , Logan RF , Devlin HB , Northfield TC . Risk assessment after acute upper gastrointestinal haemorrhage . Gut . 1996 ; 38 ( 3 ): 316 - 321 .
36. Vreeburg EM , Terwee CB , Snel P , et al. Validation of the Rockall risk scoring system in upper gastrointestinal bleeding . Gut . 1999 ; 44 ( 3 ): 331 - 335 .
37. Kollef MH , Shorr A , Tabak YP , Gupta V , Liu LZ , Johannes RS . Epidemiology and outcomes of health-care-associated pneumonia: results from a large US database of culture-positive pneumonia . Chest . 2005 ; 128 ( 6 ): 3854 - 3862 . doi: 10 .1378/chest.128.6.3854.
38. Oake N , Taljaard M , van Walraven C , Wilson K , Roth V , Forster AJ . The effect of hospital-acquired Clostridium difficile infection on in-hospital mortality . Arch Intern Med . 2010 ; 170 ( 20 ): 1804 - 1810 . doi: 10 .1001/archinternmed. 2010 . 405 .
39. Miller M , Gravel D , Mulvey M , et al. Health care-associated Clostridium difficile infection in Canada: patient age and infecting strain type are highly predictive of severe outcome and mortality . Clin Infect Dis . 2010 ; 50 ( 2 ): 194 - 201 . doi: 10 .1086/649213.
40. Dubberke ER , Butler AM , Reske KA , et al. Attributable outcomes of endemic Clostridium difficile-associated disease in nonsurgical patients . Emerging Infect Dis . 2008 ; 14 ( 7 ): 1031 - 1038 . doi: 10 .3201/eid1407. 070867 .
41. Kwok CS , Arthur AK , Anibueze CI , Singh S , Cavallazzi R , Loke YK . Risk of Clostridium difficile infection with acid suppressing drugs and antibiotics: meta-analysis . Am J Gastroenterol . 2012 ; 107 ( 7 ): 1011 - 1019 . doi: 10 .1038/ajg. 2012 . 108 .
42. Janarthanan S , Ditah I , Adler DG , Ehrinpreis MN . Clostridium difficileassociated diarrhea and proton pump inhibitor therapy: a meta-analysis . Am J Gastroenterol . 2012 ; 107 ( 7 ): 1001 - 1010 . doi: 10 .1038/ajg. 2012 . 179 .
43. Gulmez SE , Holm A , Frederiksen H , Jensen TG , Pedersen C , Hallas J . Use of proton pump inhibitors and the risk of community-acquired pneumonia: a population-based case-control study . Arch Intern Med . 2007 ; 167 ( 9 ): 950 - 955 . doi: 10 .1001/archinte.167.9.950.
44. Herzig SJ , Rothberg MB , Feinbloom DB , et al. Risk factors for nosocomial gastrointestinal bleeding and use of acid-suppressive medication in non-critically ill patients . J Gen Intern Med . 2013 . doi: 10 .1007/ s11606-012-2296-x.
45. Reimer C , Søndergaard B , Hilsted L , Bytzer P . Proton-pump inhibitor therapy induces acid-related symptoms in healthy volunteers after withdrawal of therapy . Gastroenterology . 2009 ; 137 ( 1 ): 80 -7- 87 . e1 . doi: 10 . 1053/j.gastro. 2009 . 03 .058.
46. Miller MA , Hyland M , Ofner Agostini M , Gourdeau M , Ishak M , Canadian Hospital Epidemiology Committee. Canadian Nosocomial Infection Surveillance Program. Morbidity, mortality, and healthcare burden of nosocomial Clostridium difficile-associated diarrhea in Canadian hospitals . Infect Control Hosp Epidemiol . 2002 ; 23 ( 3 ): 137 - 140 . doi: 10 .1086/502023.