This prospective interventional study was conducted over a 10-month period from June 2024 to March 2025. A trained pharmacist was stationed within the Pediatric Medical Ward (PMW) and Pediatric High-Dependency Unit (PHDU) of a tertiary care public hospital in Karachi, Pakistan, during the morning shift (0900–1500). Ethical approval for this study was obtained from the Institutional Review Board of Dow University of Health Sciences (Ref: IRB-3464/DUHS/Approval/2024/131). This study was conducted in accordance with the Declaration of Helsinki, a written informed consent in both English and Urdu (local language) was obtained from the parents or legal guardians. Additionally written assent was obtained from children age 7 years and above.

Children between 2 months and 14 years who were admitted to PMW or PHDU with a confirmed diagnosis of one or more CAIs were eligible for inclusion in the study. The CAIs diagnosis included enteric fever, acute gastroenteritis, pneumonia, dengue fever, malaria, meningitis, urinary tract infections, viral hepatitis, bronchitis, and skin and soft tissue infections. Patients admitted to the Intensive Care Unit (ICU) or Neonatal Intensive Care Unit (NICU) were excluded from the study. Patients with hospital stay less than 24 h (early discharge) were also excluded to ensure that patients received at least one full day of pharmacotherapy, allowing adequate time for medication order review and DRP identification. The patients with incomplete medical records, or those whose medical records were not available due to any reason were also excluded from the study.

OpenEpi (version 3.01) was used to calculate the sample size for a single population proportion. Assuming 50% anticipated frequency of DRPs, with a 5% margin of error and 95% confidence interval, the minimum statistically significant sample requirement was 382 patients [25]. To account for the potential attrition (incomplete records, early discharges), we enrolled 400 patients, providing 80% power to detect a 10% difference in error rates between subgroups with α = 0.05.

Data collection utilized a structured form (Supplementary Material 1), which was developed based on previous studies of pediatric DRPs [26, 27]. Data were extracted from electronic medical record and patient files. The form included demographic information (Age, gender, weight, and residential area), clinical characteristics (presenting complaints of the patient, diagnosis, comorbidities, vaccination status, allergy history), laboratory data (Total leukocyte count, Serum Creatinine, liver profile, microbiology culture results etc.), medication profile of the patient (all prescribed drugs, name, dose, frequency, route, duration and indication), and hospitalization data (Length of stay and ward type etc).

DRPs were classified using the PCNE classification V9.1 [22]. The PCNE framework comprises five domains: Problem (P), which describes the nature of the DRP (e.g., treatment effectiveness, adverse events); Causes (C), which identifies the underlying reason for the DRP (e.g., dose selection, drug selection); Interventions (I), which documents the actions taken to address the DRP; Acceptance (A), which records the prescriber’s response to the intervention; and Outcome (O), which indicates the resolution status of the problem [3]. This classification enables systematic analysis and comparison of data internationally. All identified DRPs were mapped to PCNE codes based on their characteristics, and multiple causes could be assigned to a single DRP where applicable. The PCNE classification used in the study is presented in Supplementary Table 1.

The PCNE V9.1 framework, not only classify errors related to prescription, additionally captures non-prescribing issues such as monitoring errors and patient non-adherence. These DRPs were identified by comparing each medication order against the reference standards: Lexicomp Online® pediatric dosing database, Sanford Guide to Antimicrobial Therapy, WHO treatment guidelines, British National Formulary for Children (BNFc), Hospital formulary and local guidelines. Weight-based dosing guidelines from Lexicomp Online® and BNFc were consistently applied as reference standards for all pediatric medication orders to identify dosing related DRPs.

The severity of error identified was assessed using the National Coordinating Council for Medication Error Reporting and Prevention (NCC MERP) criteria [28]. The errors were classified as: Minor when an error is unlikely to cause serious harm, permanent damage, or death to the patient (e.g., minor timing issues, incomplete documentation); Moderate when an error is with a potential to require increased monitoring of the patient, or an intervention or prolonged the stay of the patient; and the severe error was an error with potential to cause serious harm, permanent damage or death to the patient.

The PCNE classification was used to identify and categorize the type and cause of DRPs, while the NCC MERP criteria were applied to assess the clinical severity (potential harm) of each identified problem. These two classification systems serve complementary roles: PCNE describes the nature and source of the problem, while NCC MERP quantifies its potential impact on patient safety.

The primary investigator was not blinded during data collection, presenting a potential source of bias. Non-blinding may have led to overestimation of DRP prevalence, as the primary investigator might have been more likely to identify DRPs knowing the study objectives. To mitigate potential bias from non-blinded assessment, all identified DRPs were verified against reference standards before classification and complex cases were discussed with an Infectious Diseases (ID) specialist for consensus. A 10% random sample of identified DRPs was independently classified by a second pharmacist using PCNE V9.1. Inter-rater reliability was assessed using Cohen’s kappa coefficient: PCNE V9.1 classification yielded κ = 0.85 (95% CI 0.78–0.92), indicating almost perfect agreement, while NCC MERP severity classification yielded κ = 0.82 (95% CI: 0.74–0.90), indicating substantial agreement [29].

The DRP rates were calculated as; DRPs per 100 prescriptions = (total DRPs/total prescriptions) × 100; DRPs per patient = total DRPs/number of patients; DRPs per 100 patient days = (total DRPs/total length of stay in days) × 100.

The pharmacist (principal investigator) was a regular hospital staff member already working in the pediatric ward. Ward rounds were already part of routine clinical practice before the study period. During this study, the pharmacist continued these routine ward rounds and prospectively collected data for the study. When a DRP was identified, the pharmacist carried out the following interventions: (1) verified the DRP against available reference standards; (2) formulated a specific recommendation based on evidence-based guidelines; (3) communicated the recommendation to the prescribing physician; (4) discussed complex antimicrobial cases with an ID specialist; and (5) documented the intervention and outcome using PCNE codes. This sequential process followed a standardized framework reproducible in other hospital settings.

Communication methods varied based on prescriber availability and urgency. The primary method was verbal discussion during daily ward rounds (0900-1500 shift), where the pharmacist and prescriber reviewed the medication order together (80.4% of prescriber-level interventions). For complex antimicrobial cases, the pharmacist first discussed the recommendation with an ID specialist, who then provided input to the primary team. When the prescriber was not available during rounds, interventions were communicated by phone to the prescriber (10.0% of prescriber-level interventions). All communication methods were documented in the PCNE coding sheet. Intervention acceptance and implementation were distinguished using PCNE V9.1 definitions: ‘accepted’ indicated that the prescriber agreed with the pharmacist’s recommendation in principle, while ‘implemented’ indicated that the recommended change was actually applied to the patient’s medication order. An intervention could be accepted but not implemented due to practical barriers (e.g., patient discharged, or clinical decision to defer change). The status of all interventions was tracked by the pharmacists and coded for: acceptance (A1-accepted, A2 – not accepted); implementation (A1.1 – accepted and fully implemented, A1.2 – accepted but partially implemented, A1.3 – accepted but not implemented, A1.4 – accepted but implementation unknown); and DRP resolution (O1 - solved, O2 - partially solved, O3 - not solved). The standardized intervention workflow is illustrated in Figure 1.

Direct cost savings were calculated for the interventions that were associated with the reduction of the medication expenses, following the pharmacoeconomic principles [30, 31]. The cost analysis comprised of categories including: wrong dose calculation including dose increase or decrease (C3); duplication of therapy (C1.4); drug formulation changes (C2.1) including IV to oral switch or formulation changes; inappropriate drug choice (C1.1); inappropriate drug combinations (C1.3) i.e., saving from discontinuation of unnecessary medications; dose frequency adjustments (C3.3, C3.4); treatment duration adjustments (C4) including shortened or extended duration of the medication. The detailed formula and categorization of cost impact analysis is presented in Supplementary Material 2. The formula used for cost calculations were; cost savings = recommended drug cost - original drug cost, while drug cost = drug acquisition cost x daily dose x duration of therapy. The cost of the drugs was taken from the standard price list of Drug Regulatory Authority of Pakistan (DRAP) 2023 price list, and the hospital computerized pharmacy records were used to verify acquisition costs. The hospital drug prices remained fixed throughout the study period and all the calculations were based on actual acquisition costs not estimates. All costs are reported in Pakistani Rupees (PKR), with 1 USD ≈280 PKR which is approximate and reflects the exchange rate at the time of study. Purchasing power parity adjustments were not applied, which may affect international comparisons. Cost per DRP prevented was calculated by dividing net cost savings by the total number of corrected DRPs.

This analysis captured only the direct medication cost savings and direct pharmacist intervention costs. Indirect cost savings associated with reduction in length of stay, prevention of adverse drug events, avoiding readmission, decreased nursing time due to dosage form change, and long-term benefits associated with antimicrobial resistance were not quantified. The reported savings serve as estimates of the true impact of pharmacist intervention.

The return on investment calculations were performed using these formulas: total pharmacist time spent on interventions: estimated 2 h/day × 300 working days = 600 h. The estimate of 2 h per day was derived from daily activity logs maintained by the pharmacist during the study period, which recorded time spent on medication order review, intervention communication, and documentation. This represents the average active intervention time per working day and excludes routine clinical duties unrelated to DRP identification; pharmacist hourly cost (junior hospital pharmacist salary PKR 40,000/month ÷ 160 h) = PKR 250/hour; total intervention cost = 600 h × PKR 250 = PKR 150,000; net saving after deduction intervention cost = cost saving–total intervention cost; return on investment = (net saving after intervention ÷ total intervention cost) x 100. Sensitivity analysis was performed to test the robustness of the economic findings. One-way sensitivity analyses varied the following parameters: (1) pharmacist hourly cost (PKR 200 and PKR 300 instead of baseline PKR 250), (2) pharmacist time spent on interventions (1.5 h and 2.5 h per day instead of baseline 2 h), and (3) drug acquisition costs (±10% variation). Results are reported as range of total cost savings, return on investment, and cost per DRP prevented under these alternative scenarios.

The primary outcome was the prevalence of DRPs, measured as DRPs per 100 prescriptions and the percentage of patients with at least one DRP. Secondary outcomes included the acceptance rate of pharmacist interventions (PCNE A domain), DRP resolution rate (PCNE O domain), independent predictors of DRPs (AOR with 95% CI), direct cost savings (total PKR, net savings, cost per DRP prevented), and return on investment ratio. Descriptive classifications (not used as independent outcomes) included the distribution of DRPs by PCNE problem (P) and cause (C) domains, and severity distribution by NCC MERP criteria (minor, moderate, severe).

To ensure data quality and minimize bias, data were double entered by two independent researchers into an Excel spreadsheet. The entries were cross-verified and discrepancies were resolved by rechecking medical records and data sheets. Missing data were handled by complete case analysis, as the proportion of missing data was less than 5% for all variables. Patients with incomplete medical records or missing key variables were excluded from the respective analysis. The verified data were entered into IBM-SPSS version 23.0. Descriptive statistics (frequencies, percentages, means ± standard deviation) were calculated for types of DRPs, PCNE classifications, severity of medication error, interventions, and cost savings. Univariate logistic regression was performed for each predictor variable to calculate crude odds ratios (OR) with 95% confidence intervals. Variables showing significance at p < 0.10 in univariate analysis were considered for entry into multivariate model. The primary outcome for logistic regression was the presence of at least one DRP (binary: 0 = no DRP, 1 = ≥1 DRP). Number of DRPs per patient was used for descriptive purposes only and not as the dependent variable in regression models. Multivariable logistic regression using backward stepwise selection was then performed to identify independent predictors of DRPs, adjusting for potential confounders. The final model included variables with p < 0.05 after stepwise elimination. The events-per-variable (EPV) criterion was satisfied with 372 events (patients with DRPs) and 7 predictor variables, yielding an EPV of 53, which exceeds the recommended minimum of 10. The adjusted odds ratios (AOR) and 95% confidence intervals were reported. A p-value of <0.05 was considered statistically significant for the results.

A total of 793 pediatric patients were assessed for eligibility, of whom 400 patients were enrolled in the study; the mean was 4.79 ± 4.18 years (range: 0.1–14.0 years) and the mean weight of 16.0 ± 10.4 kg (range: 4.0–55.0 kg). Males comprised 64.5% (n = 258) of the cohort while females were 35.5% (n = 142). A total of 29.5% (n = 118) had a diagnosis of enteric fever, followed by 21.3% (n = 85) with gastroenteritis, 12% (n = 48) with pneumonia, and 10.5% (n = 42) with dengue fever. Most patients 73.5% (n = 294) were receiving more than five medications, 25.5% (n = 102) had 4 to 5 medications prescribed. A total of 65.5% (n = 262) spent an average of 7–9 days in the hospital, while 96% (n = 384) patients were febrile at the time of admission. Table 1 presents baseline characteristics.

A total of 3,842 medication orders were reviewed across 400 patients (mean of 9.6 orders per patient). A total of 2,010 DRPs were identified, yielding 52.3 DRPs per 100 prescriptions (95% CI: 50.7–53.9), and an average of 5.03 DRPs per patient (range: 0–15). The primary metric was DRPs per 100 prescriptions, as this allows comparison with published literature. DRPs per patient and DRPs per 100 patient-days are presented as secondary metrics. Of the 400 enrolled patients, 372 (93%) had at least one DRP and 28 (7%) had DRP-free prescriptions. The detailed patient flow is illustrated in Figure 2.

Table 2 presents the distribution of DRPs according to PCNE classification. Treatment effectiveness problems (P1) accounted for 86.7% (n = 1,742), with 53.4% showing no effect of treatment (P1.1) and 43.2% due to suboptimal drug effect (P1.2). Treatment safety problems (P2) constituted 12.4% (n = 250). The major cause of DRPs was dose selection errors (C3: 58.13%, n = 1,208), mainly due to under-dosing (C3.1: 40.2%) and over-dosing (C3.2: 43.7%). Inappropriate drug form (C2.1) was observed in 15.01% (n = 312) of prescriptions, followed by 13.08% (n = 278) errors related to drug selection (C1), with 52.3% due to excessive polypharmacy (C1.6). Monitoring errors (C9.1) accounted for 9.1% (n = 169), treatment duration errors (C4) were 0.48% (n = 10), and dispensing errors (C5) and drug use process (C6) errors were minimal (0.09% each). Figure 3 illustrates the distribution of PCNE cause domains.

Dose-related DRPs comprised 58.13% of overall DRPs. Table 3 provides the distribution of dose-related DRPs by the drug class. Antibiotics accounted for 67.3% (n = 813) of all dosing related DRPs, followed by supportive medications (antipyretics/analgesics) at 18.0% (n = 218), and antiemetics/others at 14.2% (n = 177).

Examples of dose-related DRPs included: ceftriaxone prescribed at 50 mg/kg/day instead of 80–100 mg/kg/day for a meningitis patient (under-dose, C3.1); paracetamol prescribed at 20 mg/kg/dose instead of 10–15 mg/kg/dose (overdose, C3.2); vancomycin prescribed without monitoring drug levels (C9.1); amoxicillin prescribed once daily instead of twice daily (C3.3).

A total of 1,986 interventions were suggested by the pharmacist for 2,010 DRPs (elaborated in Table 4). Drug-level interventions (I3) were 65.6% (n = 1,302), of which 53.8% were dose adjustments, 26.9% were formulation changes, and 9.4% were related to drug discontinuation. Prescriber-level interventions (I1) constituted 33.3% (n = 662); of these, 80.4% were discussed directly with the prescriber during ward rounds, and in 10.0% of cases interventions were proposed to the prescriber through written note in the medical record. Patient-level interventions (I2) were limited to 0.9% (n = 18), mainly involving referral of patients to the prescriber.

Of the 2,010 identified DRPs, 1,986 (98.8%) received pharmacist interventions. The remaining 24 DRPs (1.2%) did not receive interventions because the DRP was already corrected by the prescriber before pharmacist review (n = 16) or the patient had been discharged before an intervention could be made (n = 8).

Acceptance (A) and resolution (O) status according to PCNE is presented in Table 5. Pharmacist interventions demonstrated 98.7% (n = 1,960) acceptance (A1), of which 89.1% were fully implemented and 9.6% were partially implemented. Only 1.3% (n = 26) of the interventions were not accepted (A2), primarily due to non-agreement with the prescriber. The interventions were monitored for their resolution status (O), 81.1% (n = 1,631) of the identified problems were totally resolved (O1.1), and 13.9% were partially resolved (O2.1), while 5.0% remained unresolved (O3), most of them due to lack of patients’ and prescribers’ compliance.

Severity assessment of the errors using NCC MERP criteria (Figure 4) classified the 2,010 errors as 52.3% (n = 1,094) minor, 38.7% (n = 777) moderate, and 9.0% (n = 184) severe. Severe errors included prescription of vancomycin without therapeutic drug monitoring (nephrotoxicity risk), co-administration of IV ceftriaxone and calcium (precipitation risk), aminoglycoside high doses without renal monitoring, contraindicated drug combinations and tenfold dosing errors in low-weight infants.

Pharmacist-led interventions were associated with total cost savings of PKR 363,184.31 (approximately USD 1,290) over 10 months, with a 23.1:1 positive to negative cost ratio, demonstrating that for every PKR 1 spent on interventions that increased costs (therapeutically necessary changes), PKR 23 were saved through DRP corrections. Cost impact analysis was performed on 1,423 interventions that had direct cost implications. The remaining 563 interventions (e.g., dose frequency adjustments without cost change, therapeutic drug recommendations, patient counseling) had no direct medication cost impact and were excluded from the cost saving calculation. These interventions are clinically important but their economic benefit is indirect (e.g., preventing adverse events, reducing length of stay) and was not quantified in this analysis. The weighted average cost savings for 1,423 interventions amounted to an average of PKR 255 per intervention. The largest savings came from wrong dose correction (C3): 50.5% (PKR 262/intervention). Duplication of therapy (C1.4) accounted for 20.8% (average PKR 1,888/intervention, highest yield), while drug formulation changes (C2.1) accounted for 15.3% (average PKR 159/intervention). The projected annual savings = PKR 363,184.31 × (12/10) = PKR 435,821.17 (approximately USD 1,556). Per intervention analysis showed duplication of therapy and inappropriate drug choice as the highest economic yield, while dose frequency adjustments showed the lowest yield as shown in Table 6.

Cost per DRP prevented was PKR 241 (USD 0.86), calculated as total net savings (PKR 363,184) divided by number of DRPs fully corrected (n = 1,631). While accounting for the pharmacist time (600 h; PKR 150,000), the estimated net savings after deducting the intervention costs was PKR 213,184, yielding a return on investment of 142%.

Supplementary Table 2 presents illustrative high-impact pharmacist interventions. A detailed list of all 1,986 interventions reveals that dose corrections resulted in PKR 183,547, the highest saving. Correction of duplicate therapy resulted in PKR 75,532 by eliminating redundant drug combinations and unnecessary costs. Formulation changes primarily IV-to-oral switches saved PKR 55,578 enabling earlier discharge and interventions linked to unnecessary drugs saved PKR 22,005. The average cost savings per patient was PKR 908 (USD 3.24).

Figure 5 illustrates the contribution of each category to total cost savings.

Sensitivity analysis confirmed the robustness of the base-case findings (Supplementary Table 3). Total cost savings remained positive across all scenarios, ranging from PKR 326,866 (drug cost −10%) to PKR 399,502 (drug cost +10%). The return on investment ranged from 94% (pharmacist time 2.5 h/day) to 223% (pharmacist time 1.5 h/day). Cost per DRP prevented ranged from PKR 79 to PKR 241 depending on pharmacist time and hourly cost assumptions.

Univariate logistic regression analysis was performed to identify factors associated with DRPs (Supplementary Table 4). In univariate analysis, length of stay >7 days (OR 3.33, 95% CI: 1.53–7.25, p = 0.003) and urban residence (OR 0.24, 95% CI: 0.09–0.64, p = 0.005) were significantly associated with DRPs. Fever at admission showed borderline significance (OR 3.31, 95% CI: 0.89–12.35, p = 0.075). Past immunization, age, gender and number of prescribed medications showed no statistically significant association in univariate analysis (p > 0.05 for all). In multivariate analysis, four independent predictors of DRPs were identified (Table 7). The model showed good fit (Hosmer-Lemeshow p = 0.55) and acceptable discrimination (AUC = 0.78). Each additional drug increased DRP risk by 32% (AOR 1.32, 95% CI: 1.18–1.48, p < 0.001). Fever at admission was associated with nearly three-fold higher risk (AOR 2.84, 95% CI: 1.12–7.21, p = 0.028). Length of stay >7 days increased risk by 76% (AOR 1.76, 95% CI: 1.21–2.56, p = 0.003). Past immunization remained protective (AOR 0.51, 95% CI: 0.35–0.74, p < 0.001). Age, gender and residence were not significant predictors in the multivariate model (p > 0.05 for all).

The graphical Abstract of the study is presented in Figure 6.