Prescription Completeness and Selected Prescribing Indicators in Rural Vietnamese Primary Care: An Uncontrolled Before-and-After Study of a Multifaceted E-prescribing Quality Improvement Intervention

ABSTRACT

Objective:

The objective of this study was to assess whether a district-led multifaceted e-prescribing quality-improvement intervention improved outpatient prescription completeness and selected prescribing indicators in rural Vietnamese primary care.

Methods:

In this uncontrolled before-and-after study, 1000 health-insurance outpatient e-prescriptions were sampled before the intervention (September–November 2022) and 1000 after it (March–May 2023) from nine public primary healthcare facilities using the Health Information System of Vietnam Posts and Telecommunications (HIS-VNPT; locally recorded as VNPT-HIS). The intervention (December 2022–February 2023) combined prescriber education, drug-information support, audit-feedback, pharmacist-prescriber cross-checking, and administrative reminders. The primary outcome was complete key dosing-instruction documentation. Secondary outcomes were other mandatory completeness fields and selected the World Health Organization/International Network for Rational Use of Drugs prescribing indicators.

Findings:

Full address documentation remained low (31/1000 [3.1%] vs. 29/1000 [2.9%]; P = 0.79). Complete key dosing instructions increased from 956/1000 (95.6%) to 974/1000 (97.4%; P = 0.03). Mean medicines per prescription did not change significantly (4.08 ± 1.535 vs. 4.20 ± 1.348; P = 0.07), and all medicines were prescribed by international nonproprietary name. Antibiotic prescribing increased from 335/1000 (33.5%) to 447/1000 (44.7%; P < 0.001), vitamin prescribing from 136/1000 (13.6%) to 190/1000 (19.0%; P = 0.001), and essential-medicine use changed only slightly (1480/4078 [36.3%] vs. 1568/4197 [37.4%]; P = 0.31).

Conclusion:

E-prescribing may provide an infrastructure for quality improvement but requires clinical decision support and pharmacy review.

INTRODUCTION

Medication-related harm remains a global patient-safety priority, and the prescription is often the first point at which preventable harm can be recognized or introduced.[1] Prescribing errors, omitted prescription information, and inadequate communication of medicine instructions have been associated with adverse drug events and avoidable downstream workload in both inpatient and ambulatory settings.[2,3]

Electronic prescribing can standardize administrative fields and reduce selected medication errors, but digitization does not by itself ensure clinically appropriate treatment choices.[4,5,6] For primary care systems, especially in resource-constrained settings, prescription completeness therefore needs to be assessed together with medicine-use indicators such as polypharmacy, antibiotic use, international nonproprietary name prescribing, injection use, vitamin use, and essential-medicine alignment.[7,8,9]

This distinction is important in Vietnam and comparable low- and middle-income countries, where primary care must manage acute infections and chronic noncommunicable diseases while antimicrobial resistance remains a major public-health concern.[10,11,12] Multifaceted interventions combining education, audit-feedback, reminders, information support, and administrative accountability are feasible for district health systems, yet their effect on routine outpatient e-prescribing quality remains underdocumented.[13,14,15,16]

In this study, the primary outcome was the proportion of prescriptions with complete key dosing instructions, defined as documentation of dose per administration, daily dose or frequency, route, and timing. Secondary outcomes were completeness of other mandatory prescription elements and selected prescribing indicators: mean number of medicines per prescription, medicines prescribed by international nonproprietary name, prescriptions containing at least one antibiotic, injection, or vitamin, and the proportion of medicines included in the national essential medicines list. We hypothesized that prescriber education, drug-information support, audit-feedback, pharmacist-prescriber cross-checking, and administrative reminders would mainly improve documentation-sensitive outcomes, especially dosing instructions, whereas medicine-use indicators would improve less or incompletely because the package did not include diagnosis-linked clinical decision support, formulary restriction, or antimicrobial stewardship algorithms.

METHODS

We conducted an uncontrolled before-and-after interventional study in nine public primary healthcare facilities in a rural district in southern Vietnam, comprising one district outpatient clinic and eight community health stations. Baseline prescriptions were collected from September to November 2022, the intervention was implemented from December 2022 to February 2023, and postintervention prescriptions were collected from March to May 2023 [Figure 1].

Figure 1.

Before-and-after study sampling and intervention flow

Eligible records were health-insurance outpatient electronic prescriptions issued through the Health Information System of Vietnam Posts and Telecommunications (HIS-VNPT; locally recorded as VNPT-HIS), the platform used by the facilities for outpatient prescriptions and insurance administration. We excluded nonhealth-insurance prescriptions, repeat prescriptions from the same patient within the same period, and postintervention visits by patients already selected at baseline. All selected records were de-identified. The analytic extract retained final selected records but not counts for all records screened, numeric exclusions by reason, or duplicate records; this limitation is shown in Figure 1 and acknowledged in the discussion.

The intervention targeted prescriber behavior. All outpatient prescribers in the participating facilities were exposed to the activities. The de-identified extract did not include prescriber identifiers, specialty, or a longitudinal prescriber roster; therefore, the number of physicians, specialty mix, and period-to-period prescriber composition could not be analyzed.

The sample size was calculated for the primary documentation outcome using the single-proportion formula n = Z²p (1 − p)/d².[17] With α =0.05 (Z = 1.96), expected completeness P = 0.7525 from local prestudy quality-monitoring data, and precision d = 0.03, the minimum was 795 prescriptions per period. We sampled 1000 prescriptions per period to improve precision and facility-level representation. Sampling was proportional to insured outpatient volume in the first quarter of 2022: 833 prescriptions from the district outpatient clinic and 167 from community health stations in each period. Within each stratum, prescriptions were selected systematically across the 3-month collection period.

Prescription completeness was assessed using a structured checklist derived from national outpatient prescribing regulations.[18,19] The assessed fields included patient identifiers and address, required information for children aged younger than 72 months when applicable, prescriber identifiers and signature, medicine strength or concentration and quantity, use of a leading zero before single-digit quantities, and dosing instructions (dose per administration, daily dose or frequency, route, and timing).

Selected medicine-use indicators followed the World Health Organization/International Network for Rational Use of Drugs (WHO/INRUD) operational definitions and the Vietnamese national essential medicines list.[20,21] Medicines were standardized by active ingredient or international nonproprietary name. Drug-class variables were classified from the local e-prescription drug list as antibiotic, injection, or vitamin; the vitamin indicator included single-ingredient vitamins, multivitamins, vitamin–mineral combinations, and vitamin-containing prescribed products recorded in HIS-VNPT/VNPT-HIS. Essential-medicine status was determined by active-ingredient matching to the national list. The anatomical therapeutic chemical/defined daily dose system was not applied because the outcomes were not dose-normalized utilization measures.

The intervention was designed as a district-led, multifaceted quality-improvement package. The communication component included establishing a drug-information team, updating and distributing prescribing regulations and reference materials, and delivering prescriber reminders on common prescription errors. The administrative component included prescription screening by prescriber, feedback at professional meetings, reports to drug and therapeutics leadership, a pharmacist-prescriber cross-check workflow, managerial reminders for repeated errors, and facility-level support for treatment protocols.

Trained healthcare personnel were two pharmacists and one district health-center staff member trained by the study team in eligibility criteria, variable definitions, de-identification, abstraction procedures, and checking rules. They abstracted data using a standardized form, which was checked against electronic prescriptions before entry.

Categorical variables were summarized as counts and percentages and compared between periods using Chi-square tests. The mean number of medicines per prescription was summarized as mean ± standard deviation and compared using an independent t-test. All P values were two-sided, with P < 0.05 considered statistically significant. P > 0.01 are reported to two decimal places, those between 0.01 and 0.001 to three decimal places, and those < 0.001 as P < 0.001, following the journal guidance. Analyses were performed using SPSS version 18.0 (IBM Corp., Armonk, NY, USA).

The study protocol was approved by an institutional biomedical ethics committee (approval No. 22.021.HV/PCT-HĐĐĐ, July 25, 2022). The study complied with the principles of the Declaration of Helsinki and used de-identified routine prescription records.

RESULTS

We analyzed 2000 outpatient electronic prescriptions: 1000 before and 1000 after the intervention. The final sampled distribution was 833 prescriptions from the district outpatient clinic and 167 prescriptions from community health stations in each period [Figure 1]. Patient sex distribution was similar between periods (male: 526/1000 [52.6%] vs. 536/1000 [53.6%]; P = 0.65). The proportion of patients aged 21–35 years decreased from 94/1000 (9.4%) to 66/1000 (6.6%; P = 0.02), whereas the proportion aged 81 years or older declined from 53/1000 (5.3%) to 35/1000 (3.5%; P = 0.05) [Table 1].

Table 1.

Characteristics of patients in the pre- and postintervention prescription samples (n=1000 each)

Characteristic	Preintervention, n (%)	Postintervention, n (%)	P
Age group (years)
≤20	123 (12.3)	152 (15.2)	0.06
21–35	94 (9.4)	66 (6.6)	0.02
36–65	520 (52.0)	522 (52.2)	0.93
66–80	210 (21.0)	225 (22.5)	0.42
≥81	53 (5.3)	35 (3.5)	0.05
Sex
Male	526 (52.6)	536 (53.6)	0.65
Female	474 (47.4)	464 (46.4)
ICD-10 diagnostic group
Infectious and parasitic diseases	83 (8.3)	36 (3.6)	<0.001
Endocrine, nutritional, and metabolic diseases	234 (23.4)	192 (19.2)	0.02
Diseases of the circulatory system	173 (17.3)	149 (14.9)	0.14
Diseases of the respiratory system	223 (22.3)	287 (28.7)	0.001
Diseases of the digestive system	48 (4.8)	52 (5.2)	0.68
Diseases of the musculoskeletal system and connective tissue	83 (8.3)	90 (9.0)	0.58
Other diagnoses	156 (15.6)	194 (19.4)	0.03

ICD-10, 10^th Revision. P values are from Chi-square tests. Significant P values=Age 21–35 years, P=0.02; infectious and parasitic diseases, P<0.001; endocrine, nutritional, and metabolic diseases, P=0.02, respiratory diseases, P=0.001; other diagnoses, P=0.03. ICD=International Classification of Diseases

The diagnostic case-mix differed between periods. Prescriptions for infectious and parasitic diseases decreased from 83/1000 (8.3%) to 36/1000 (3.6%; P < 0.001), whereas prescriptions for respiratory diseases increased from 223/1000 (22.3%) to 287/1000 (28.7%; P = 0.001). Endocrine, nutritional, and metabolic diseases decreased from 234/1000 (23.4%) to 192/1000 (19.2%; P = 0.02) [Table 1].

Most mandatory administrative and medicine information fields were complete in both periods [Table 2]. Patient name, age, sex, prescription date, prescriber name and professional title, prescriber signature, medicine strength or concentration, medicine quantity, and leading-zero notation for single-digit quantities were complete in all selected prescriptions. Full patient address was rarely documented, with no meaningful change after the intervention (31/1000 [3.1%] vs. 29/1000 [2.9%]; P = 0.79). The primary outcome, complete key dosing-instruction documentation, increased from 956/1000 (95.6%) to 974/1000 (97.4%; P = 0.03).

Table 2.

Completeness of key outpatient e-prescription elements before and after the intervention (n=1000 each)

Element	Preintervention, n (%)	Postintervention, n (%)	P
Patient information
Patient name	1000 (100.0)	1000 (100.0)	NA
Age	1000 (100.0)	1000 (100.0)	NA
Sex	1000 (100.0)	1000 (100.0)	NA
Full address	31 (3.1)	29 (2.9)	0.79
Required information for children aged <72 months (or not applicable)	999 (99.9)	998 (99.8)	0.56
Prescriber information
Prescription date	1000 (100.0)	1000 (100.0)	NA
Prescriber name and professional title	1000 (100.0)	1000 (100.0)	NA
Prescriber signature	1000 (100.0)	1000 (100.0)	NA
Medicine information
Strength or concentration	1000 (100.0)	1000 (100.0)	NA
Quantity	1000 (100.0)	1000 (100.0)	NA
Leading zero before single-digit quantities	1000 (100.0)	1000 (100.0)	NA
Dosing instructions
Dose per administration	956 (95.6)	974 (97.4)	0.03
Daily dose or frequency	956 (95.6)	974 (97.4)	0.03
Route of administration	1000 (100.0)	1000 (100.0)	NA
Timing of administration	956 (95.6)	974 (97.4)	0.03

Because no between-period test was needed when both proportions were 100%. P values are from Chi-square tests. Significant P values=Dose per administration, daily dose or frequency, and timing of administration, each P=0.03. NA=Not applicable

The mean number of medicines per prescription did not change significantly (4.08 ± 1.535 vs. 4.20 ± 1.348; P = 0.07), and all medicines were prescribed by international nonproprietary name in both periods [Table 3]. Prescriptions containing an antibiotic increased from 335/1000 (33.5%) to 447/1000 (44.7%; P < 0.001), and prescriptions containing a vitamin increased from 136/1000 (13.6%) to 190/1000 (19.0%; P = 0.001). Injection prescribing was rare (one prescription before the intervention and none after; P = 0.32). The proportion of medicines included in the national essential medicines list changed only slightly, from 1480/4,078 (36.3%) to 1568/4197 (37.4%; P = 0.31).

Table 3.

Selected prescribing indicators before and after the intervention

Indicator	Preintervention	Postintervention	P
Mean number of medicines per prescription, mean±SD	4.08±1.535	4.20±1.348	0.07
Medicines prescribed by INN, n/N (%)	4078/4078 (100.0)	4197/4197 (100.0)	NA
Prescriptions containing ≥1 antibiotic, n/N (%)	335/1000 (33.5)	447/1000 (44.7)	<0.001
Prescriptions containing ≥1 injection, n/N (%)	1/1000 (0.1)	0/1000 (0.0)	0.32
Prescriptions containing ≥1 vitamin, n/N (%)	136/1000 (13.6)	190/1000 (19.0)	0.001
Medicines included in the national essential medicines list, n/N (%)	1480/4078 (36.3)	1568/4197 (37.4)	0.31

P value for the mean number of medicines is from an independent t-test; all other P values are from Chi-square tests. Percentages for INN and essential-medicine indicators use the number of medicines as the denominator; antibiotic, injection, and vitamin indicators use the number of prescriptions as the denominator. Significant P values: antibiotic prescribing, P<0.001, vitamin prescribing, P=0.001. INN=International nonproprietary name, NA=Not applicable, SD=Standard deviation

DISCUSSION

This uncontrolled before-and-after study supports the practical interpretation that e-prescribing can provide infrastructure for prescription-quality improvement, but it is insufficient as a stand-alone intervention. Most mandatory fields were already highly complete, full address documentation remained poor, complete key dosing-instruction documentation improved modestly, and selected prescribing indicators did not improve.

The high completeness of patient identifiers, prescriber information, medicine strength, and medicine quantity is consistent with e-prescribing systems that standardize data entry and reduce omissions common in handwritten prescriptions.[4,5,6] However, low full-address documentation indicates that electronic systems can also normalize omissions if required fields are not technically enforced. Education alone is unlikely to solve fields controlled by software design or registration workflow.

The improvement in dosing instructions is clinically relevant despite its modest absolute size. Dose, daily frequency, route, and timing are the components most directly interpreted by patients, pharmacists, and nurses. Missing or ambiguous instructions may increase clarification workload, reduce adherence, and create risk when patients receive multiple medicines. The residual gap suggests that automated prompts or hard-stop fields should be considered.

The prescribing indicators need to be interpreted as screening indicators rather than direct judgments of appropriateness. The mean number of medicines was approximately four per prescription in both periods, which is higher than the 3.52 medicines per prescription reported in Chinese county hospitals using WHO/INRUD indicators.[22] Polypharmacy can be appropriate for multimorbidity, but without medication review, it increases the risk of drug–drug interactions, adverse events, nonadherence, and avoidable cost.[9] The essential-medicine indicator changed only slightly, suggesting that the intervention did not substantially alter formulary alignment. Essential-medicine policies can improve access and equity when they are connected to procurement, reimbursement, prescribing protocols, and decision support, but passive availability of an essential list is usually insufficient.[8]

The increase in antibiotic prescribing deserves careful interpretation. Our all-diagnosis antibiotic-prescription rate rose from 33.5% to 44.7%, whereas Vietnamese primary care data for acute respiratory infections reported antibiotic prescribing in 97.0% of 193,010 visits.[23] These denominators differ, but the increase remains concerning because the postintervention period had more respiratory diagnoses, and Vietnam has persistent problems with antibiotic access, community use, and resistance.[10,11,12] Future interventions should add diagnosis-linked antibiotic feedback, local algorithms, and clinical pharmacist review.

Vitamin prescribing increased from 13.6% to 19.0%. The indicator included single-ingredient vitamins, multivitamins, vitamin-mineral combinations, and vitamin-containing prescribed products. Because diagnoses were available only at International Classification of Diseases-10 chapter level, individual vitamin prescriptions could not be linked to deficiency diagnoses, laboratory results, pregnancy status, nutritional assessment, or other patient-level indications. We therefore did not classify individual vitamin prescriptions as appropriate or inappropriate; the increase should be interpreted as a signal for diagnosis-linked medication review.

The findings help explain why multifaceted interventions may improve documentation but not therapeutic choices. Audit feedback is more effective when feedback is timely, individualized, repeated, and paired with clear action targets.[13] Ambulatory antibiotic-prescribing interventions often require accountable justification, peer comparison, explicit commitment, or clinical decision support.[13,14,15,16,24] In rural primary care, HIS-VNPT/VNPT-HIS could flag incomplete instructions and high-risk patterns, whereas pharmacists focus review on antibiotics, vitamins, polypharmacy, essential-medicine alignment, and drug–drug interaction risk.

Strengths include routinely generated e-prescription data from multiple primary care facilities, systematic sampling across two periods, and evaluation of a pragmatic district-level intervention. The analysis separated prescription completeness from medicine-use indicators, recognizing that an administratively complete prescription can still be clinically suboptimal.

Limitations include the uncontrolled design, secular trends, seasonal variation, and case-mix changes. The analytic extract did not retain counts for all eligible prescriptions before sampling, numeric exclusions by reason, duplicate records, or prescriber-level identifiers; therefore, prescriber composition and clustering could not be evaluated. We assessed selected indicators rather than diagnosis-specific appropriateness and did not evaluate patient outcomes, drug–drug interactions, adherence, costs, or pharmacist intervention acceptance. Generalizability beyond one rural district may be limited.

CONCLUSION

E-prescribing may provide an infrastructure for quality improvement but requires clinical decision support and pharmacy review.

Author contributions

Hieu Tri Bui: Investigation, data curation, formal analysis, methodology, visualization, and writing – original draft. Tu Thi Cam Le: Methodology, validation, supervision, and writing – review and editing. Le Thi Nhan Duyen: Project administration, institutional coordination, resources, validation, and writing – review and editing. Thang Nguyen: Conceptualization, methodology, supervision, project administration, validation, and writing – review and editing. All authors read and approved the final version of the manuscript and agreed to be accountable for all aspects of the work.

Data availability

De-identified data may be made available upon reasonable request to the corresponding author, subject to institutional policies.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.