Table 2 Methodology and results of the retrospective studies included in a literature review of drug sales data analyses for surveillance of infectious diseases
Author | Syndrome | Reference data sources | Drugs selected | Statistical methods | Correlation strength | Correlation timeliness | Detection sensitivity | Detection specificity | Detection timeliness |
|---|---|---|---|---|---|---|---|---|---|
STUDIES ON PRESCRIBED DRUGS | |||||||||
Sugawara et al. [27] 2012 | ILI | Influenza cases from 5000 hospitals and clinics | Drugs against influenza virus : oseltamivir, zanamivir, laninamivir | Correlation | Pearson corr. coeff. r = 0.992 for 2009/10 and r = 0.972 for 2010/11 (p<0.001) | - | - | - | - |
Polgreen et al. [23] 2011 | GI | Hospitalizations with diagnosis of Clostridium Difficile Infections | Oral vancomycin | Cross-correlation, Regression model | Increase in Clostridium Difficile Infections associated with increase in vancomycin use | - | - | - | - |
Chopra et al. [12] 2011 | ILI | Cases of influenza reported from nine sentinel healthcare providers | Oseltamivir | Correlation | Spearman corr.coeff. r = 0.46 (p<0.003) Peaks occurred at the same time | - | - | - | - |
Yoshida et al. [30] 2009 | ILI | 28 sentinel surveillance sites of influenza in Sahai City (clinics and hospitals) | Oseltamivir and Zanamivir | Correlation | Pearson corr.coeff. r = 0.954 | - | - | - | - |
Van den Wijngaard et al. [28] 2008 | ILI | Respiratory pathogen diagnosis in laboratory registries (Influenza A, B, RSV, enterovirus, S.pneumoniae..) | Drugs for respiratory infectious diseases (7 ATC classes) | Graphical comparison, Correlation, Linear regression model | Pearson corr .coeff. r = 0.60 for Influenza A, r = 0.58 for RSV, r = 0.60 for S. pneumonia, r = 0.39 for influenza B (p<0.05) 80% of variation explained by respiratory pathogens | 2 weeks earlier until 1 week later | - | - | - |
Author | Syndrome | Reference data sources | Drugs selected | Statistical methods | Correlation strength | Correlation timeliness | Detection sensitivity | Detection specificity | Detection timeliness |
Chen et al. [11] 2005 | Pertussis | Reported cases of pertussis from the NYS department of Health | Macrolide antibiotics | CUSUM | - | - | 100% The signal was indicator of pertussis outbreak | 100% | Not early warning |
Couturier et al. [13] 2004 | Syphilis | Reported cases of syphilis from hospitals, physicians, sexually transmitted disease clinics. | Benzylpenicillin benzathine 2.4 MUI | Descriptive analysis | Similar trend (+22% increase in Paris, +10% in the 5 regions) | Similar trend | - | - | - |
STUDIES ON OTC DRUGS | |||||||||
Kirian et al. [19] 2010 | GI | Cases of gastrointestinal diseases from County Health Department and detected GI outbreaks | Diarrheal remedies (based on common use) | Cross-correlation, Regression ARIMA | No significant correlation between sales and GI cases counts, outbreak counts. | - | Not sensitive 4%-14% | Specific 97%-100% | - |
Edge et al. [17] 2006 | GI | Counts of GI cases due to bacteria, parasites, and viruses | Antinauseant and antidiarrheal products | Cross-correlation | Temporal patterns of OTC and Norovirus activity were similar Pearson r2 = 0.44 | No delay | - | - | - |
Okhusa et al. [22] 2005 | ILI | Reporting of patients with ILI (hospitals, clinics, physicians) | Most common treatments ILI | Cross-correlation, Prediction model, Peak comparison | Significant correlation between sales and influenza activity | Sales do not determine influenza in advance | - | - | - |
Das et al. [14] 2005 | ILI | Emergency department in New York City (ratio of ILI syndrome visits/other syndrome visits) | A cold medication selected statistically from a group of 400 cold medications (ratio ILI/analgesics drugs sales) | Cross-correlation, Serfling method, Graphical comparison | High correlation Pearson r2 = 0.60 (p<0.001) | No lead time | Sensitive (data not reported) | Not specific (not reported) | Not earlier warning than reference data |
Das et al. [14] 2005 | GI | Emergency department in New York City (ratio of GI syndrome visits/other syndrome visits) | Common antidiarrheal drugs(ratio GI/analgesic drug sales) | Cross-correlation, Graphical comparison | Low correlation Pearson r2 = 0.24 (p<0.005) Similar increases during the fall (norovirus) and influenza peak. Increase in ED GI visits during late winter (rotavirus), but no increase in drug sales. | - | Less sensitive than ED system | - | - |
Author | Syndrome | Reference data sources | Drugs selected | Statistical methods | Correlation strength | Correlation timeliness | Detection sensitivity | Detection specificity | Detection timeliness |
Edge et al. [16] 2004 | GI | Emergency room visits for acute GI, number of GI cases from case series investigations (waterborne outbreak) | Saskatchewan: four commonly used antidiarrheals and antinauseants Ontario: 12 products (antidiarrheal, antinauseant, rehydration products) | Graphical comparison(Ontario, Saskatchewan), CUSUM, moving average (Ontario) | Trends of OTC products comparable to the outbreak epidemic curve (Saskatchewan,Ontario) | - | 100% exceeded threshold during the outbreak period (Ontario) | 100% | Not earlier |
Magruder et al. [21] 2004 | ILI | Outpatient insurance-claim diagnoses for acute respiratory conditions, from 13,000 clinics and doctors’ offices | Remedies for treating influenza (common use) | Cross-correlation | Seasonal trend: Pearson r (between 0.95 and 0.99) | 1- 3 week lead | - | - | - |
Non-seasonal trend: Pearson r (between 0.25 and 0.75) | No repeatable lead time | ||||||||
Hogan et al. [18] 2003 | ILI and GI | Hospital-discharge diagnoses of respiratory and diarrheal disease in children (for all hospitals in Pennsylvania, in Utah, and 95% of Indiana). | Electrolyte products | Cross-correlation, EWMA | Pearson r = 0.90 (95% CI, 0.87-0.93) | Sales preceded diagnoses by 1.7 weeks (95% CI, 0.50-2.9) | 100% | 100% | Electrolyte sales preceded detection from diagnoses by an average of 2.4 weeks (95% CI, 0.1-4.8) Detection earlier in 12/18 outbreaks |
Magruder [20] 2003 | ILI | Outpatient insurance-claim diagnoses for acute respiratory conditions | Cold remedies: 622 products (then grouped in categories by an expert in pharmacoepidemiology) | Cross-correlation | Pearson r = 0.9 | Mean lead times of 2.8 days | - | - | - |
Davies et al. [15] 2003 | ILI | Emergency admission data from Nottingham City Hospital NHS Trust. | Cold and flu remedies (cold, cough, decongestant, throat preparation) | Correlation, Peak comparison, Threshold detection method | National and local sales positively correlated with admissions in 98/99 and 99/00, not 00/01 | - | 100%(for local sales) | 100% (for local sales) | Rate of local sales exceed threshold of 1000 units per week 2 weeks prior to peak in emergency admissions |
Author | Syndrome | Reference data sources | Drugs selected | Statistical methods | Correlation strength | Correlation timeliness | Detection sensitivity | Detection specificity | Detection timeliness |
Stirling et al. [26] 2001 | GI | Telephone survey from a sample of households: number of persons with diarrheal symptoms and/or with stool specimen positive to C. parvum oocysts. (waterborne outbreak) | Common antidiarrheal (determined by each pharmacist) | Descriptive analysis | A fivefold increase in sales during the epidemic period | - | - | - | - |
Proctor et al. [24] 1998 | GI | Comparison with eight sources (laboratory confirmed cases of Cryptosporidium, clinically defined cases) (waterborne outbreak) | Antidiarrheal: Imodium, Pepto Bismol, Kaopectate | Descriptive analysis | Significant increase in drug sales during epidemic period | - | - | - | - |
Rodman et al. [25] 1997 | GI | Cases of cryptosporidiosis (5 waterborne outbreaks) | Antidiarrheal drugs | Descriptive analysis | Milkauwee: increased 20 fold; Las Vegas: no data; Collingwood: increased in 2 of 3 stores;Kelowna: increased 3 fold;Cranbrook: increased | - | - | - | - |
Welliver et al. [29] 1979 | ILI | Laboratory count of influenza B | Children’s aspirin, adult antipyretics, cold remedies | Determination of the% of sales increase, peak comparison | Sales of cold remedies averaged 185% above the baseline value during the peak influenza activity | - | - | - | - |