| |
| """ |
| Laboratory Information Systems Dataset |
| ========================================== |
| Each record = ONE laboratory facility's LIS status/metrics. |
| |
| Literature-Grounded Parameterization: |
| [1] Nguyen TT et al. (2018). LIS in low-resource settings. |
| BMC Med Inform Decis Mak. DOI: 10.1186/s12911-018-0693-x |
| - LIS adoption SSA: <15% of labs |
| - Paper-based systems: >80% of SSA labs |
| - Barriers: cost, infrastructure, training, internet |
| |
| [2] Were MC et al. (2015). Implementation of electronic health |
| records in SSA. JAMIA. DOI: 10.1093/jamia/ocu023 |
| - Open-source LIS: OpenELIS, BLIS, iLab |
| - Internet connectivity: <30% of health facilities in SSA |
| - Power supply: intermittent in 40-60% of facilities |
| |
| [3] Sarkinfada F et al. (2012). Lab informatics in Nigeria. |
| Afr J Lab Med. DOI: 10.4102/ajlm.v1i1.18 |
| - Data quality: 30-50% of paper records incomplete |
| - TAT tracking: only with LIS |
| - Result delivery: SMS, paper, phone |
| """ |
| import numpy as np, pandas as pd, argparse, os |
|
|
| SCENARIOS = { |
| 'lis_implemented': { |
| 'exemplar': 'South Africa/Kenya/Nigeria (urban)', |
| 'lis_available': 0.45, |
| 'lis_type_electronic': 0.35, |
| 'paper_register': 0.55, |
| 'internet_available': 0.55, |
| 'power_reliable': 0.60, |
| 'barcode_scanning': 0.20, |
| 'auto_result_entry': 0.15, |
| 'data_completeness': 0.75, |
| 'tat_tracked': 0.40, |
| 'result_sms_delivery': 0.25, |
| 'dhis2_reporting': 0.45, |
| 'data_backup': 0.35, |
| 'staff_trained_lis': 0.30, |
| 'interoperability': 0.10, |
| }, |
| 'partial_lis': { |
| 'exemplar': 'Kenya (district)/Ghana/Tanzania/Ethiopia', |
| 'lis_available': 0.10, |
| 'lis_type_electronic': 0.08, |
| 'paper_register': 0.85, |
| 'internet_available': 0.20, |
| 'power_reliable': 0.30, |
| 'barcode_scanning': 0.03, |
| 'auto_result_entry': 0.02, |
| 'data_completeness': 0.50, |
| 'tat_tracked': 0.10, |
| 'result_sms_delivery': 0.08, |
| 'dhis2_reporting': 0.25, |
| 'data_backup': 0.08, |
| 'staff_trained_lis': 0.05, |
| 'interoperability': 0.02, |
| }, |
| 'no_lis': { |
| 'exemplar': 'DRC/CAR/Sierra Leone/Niger (rural)', |
| 'lis_available': 0.02, |
| 'lis_type_electronic': 0.01, |
| 'paper_register': 0.95, |
| 'internet_available': 0.05, |
| 'power_reliable': 0.10, |
| 'barcode_scanning': 0.005, |
| 'auto_result_entry': 0.005, |
| 'data_completeness': 0.30, |
| 'tat_tracked': 0.02, |
| 'result_sms_delivery': 0.02, |
| 'dhis2_reporting': 0.08, |
| 'data_backup': 0.02, |
| 'staff_trained_lis': 0.01, |
| 'interoperability': 0.005, |
| }, |
| } |
|
|
| def generate_dataset(n=10000, seed=42, scenario='partial_lis'): |
| rng = np.random.default_rng(seed) |
| sc = SCENARIOS[scenario] |
| records = [] |
| for idx in range(n): |
| rec = {'id': idx + 1} |
| rec['facility_level'] = rng.choice(['national_reference','regional','district','health_centre','private'], |
| p=[0.05, 0.12, 0.28, 0.40, 0.15]) |
| rec['location'] = rng.choice(['urban','peri_urban','rural','remote'], p=[0.25, 0.20, 0.30, 0.25]) |
| rec['ownership'] = rng.choice(['government','ngo','private','faith_based'], p=[0.55, 0.12, 0.20, 0.13]) |
|
|
| tier = 1.5 if rec['facility_level'] in ['national_reference','regional','private'] else 0.6 |
| urban_boost = 1.3 if rec['location'] in ['urban','peri_urban'] else 0.7 |
|
|
| |
| rec['power_reliable'] = 1 if rng.random() < sc['power_reliable']*urban_boost else 0 |
| rec['solar_backup'] = 1 if (not rec['power_reliable'] and rng.random() < 0.15) else 0 |
| rec['generator_available'] = 1 if rng.random() < sc['power_reliable']*0.8 else 0 |
| rec['internet_available'] = 1 if rng.random() < sc['internet_available']*urban_boost else 0 |
| if rec['internet_available']: |
| ip = np.array([0.15*tier, 0.40, 0.05, max(0.01, 0.40-0.15*tier)]) |
| ip = ip / ip.sum() |
| rec['internet_type'] = rng.choice(['broadband','mobile_data','satellite','none'], p=ip) |
| else: |
| rec['internet_type'] = 'none' |
|
|
| rec['computers_available'] = max(0, int(rng.poisson(3*tier))) if rec['power_reliable'] else 0 |
|
|
| |
| rec['lis_available'] = 1 if rng.random() < min(0.90, sc['lis_available']*tier*urban_boost) else 0 |
| if rec['lis_available']: |
| rec['lis_type'] = rng.choice(['openellis','blis','commercial','custom_built','spreadsheet'], |
| p=[0.20, 0.15, 0.25, 0.15, 0.25]) |
| rec['lis_fully_electronic'] = 1 if rng.random() < sc['lis_type_electronic']*tier else 0 |
| else: |
| rec['lis_type'] = 'none' |
| rec['lis_fully_electronic'] = 0 |
| rec['paper_register'] = 1 if rng.random() < sc['paper_register'] else 0 |
| rec['dual_system'] = 1 if (rec['lis_available'] and rec['paper_register']) else 0 |
|
|
| |
| rec['barcode_scanning'] = 1 if (rec['lis_available'] and rng.random() < sc['barcode_scanning']*tier) else 0 |
| rec['auto_result_entry'] = 1 if (rec['lis_available'] and rng.random() < sc['auto_result_entry']*tier) else 0 |
| rec['order_entry_electronic'] = 1 if (rec['lis_available'] and rng.random() < sc['lis_type_electronic']*1.2) else 0 |
| rec['result_sms_delivery'] = 1 if rng.random() < sc['result_sms_delivery'] else 0 |
| rec['tat_tracked'] = 1 if rng.random() < sc['tat_tracked'] else 0 |
| rec['stock_management_electronic'] = 1 if (rec['lis_available'] and rng.random() < 0.20) else 0 |
| rec['quality_module'] = 1 if (rec['lis_available'] and rng.random() < 0.15) else 0 |
|
|
| |
| rec['data_completeness_pct'] = int(np.clip(rng.normal(sc['data_completeness']*100, 15), 10, 100)) |
| rec['data_timeliness_pct'] = int(np.clip(rng.normal(sc['data_completeness']*80, 20), 5, 100)) |
| rec['data_accuracy_pct'] = int(np.clip(rng.normal(sc['data_completeness']*90, 12), 20, 100)) |
| rec['duplicate_records_pct'] = max(0, int(rng.normal(100-sc['data_completeness']*100, 8))) |
|
|
| |
| rec['dhis2_reporting'] = 1 if rng.random() < sc['dhis2_reporting'] else 0 |
| rec['monthly_report_timely'] = 1 if rng.random() < sc['data_completeness']*0.8 else 0 |
| rec['aggregate_only'] = 1 if (rec['dhis2_reporting'] and not rec['lis_available']) else 0 |
| rec['patient_level_data'] = 1 if (rec['lis_available'] and rng.random() < 0.40) else 0 |
|
|
| |
| rec['data_backup'] = 1 if rng.random() < sc['data_backup'] else 0 |
| rec['backup_offsite'] = 1 if (rec['data_backup'] and rng.random() < 0.30) else 0 |
| rec['password_protected'] = 1 if (rec['lis_available'] and rng.random() < 0.60) else 0 |
| rec['it_support_available'] = 1 if rng.random() < sc['staff_trained_lis']*1.5 else 0 |
| rec['staff_trained_lis'] = 1 if rng.random() < sc['staff_trained_lis'] else 0 |
| rec['interoperability_emr'] = 1 if rng.random() < sc['interoperability'] else 0 |
|
|
| |
| rec['tests_per_month'] = max(10, int(rng.lognormal(np.log(500*tier), 0.8))) |
| rec['test_menu_items'] = max(3, int(rng.normal(20*tier, 8))) |
|
|
| rec['year'] = rng.choice([2019,2020,2021,2022,2023], p=[0.12,0.18,0.20,0.25,0.25]) |
| records.append(rec) |
|
|
| df = pd.DataFrame(records) |
| print(f"\n{'='*60}\nLaboratory Information Systems — {scenario} ({sc['exemplar']})") |
| print(f" LIS available: {df['lis_available'].mean()*100:.1f}% | Electronic: {df['lis_fully_electronic'].mean()*100:.1f}%") |
| print(f" Internet: {df['internet_available'].mean()*100:.1f}% | Power: {df['power_reliable'].mean()*100:.1f}%") |
| print(f" Data completeness: {df['data_completeness_pct'].mean():.0f}% | DHIS2: {df['dhis2_reporting'].mean()*100:.1f}%") |
| return df |
|
|
| if __name__ == '__main__': |
| parser = argparse.ArgumentParser() |
| parser.add_argument('--all-scenarios', action='store_true') |
| parser.add_argument('--n', type=int, default=10000) |
| parser.add_argument('--seed', type=int, default=42) |
| args = parser.parse_args() |
| os.makedirs('data', exist_ok=True) |
| if args.all_scenarios: |
| for sc in SCENARIOS: |
| df = generate_dataset(n=args.n, seed=args.seed, scenario=sc) |
| df.to_csv(os.path.join('data', f'lis_{sc}.csv'), index=False) |
| print(f" -> Saved\n") |
| else: |
| df = generate_dataset(n=args.n, seed=args.seed) |
| df.to_csv(os.path.join('data', 'lis_partial_lis.csv'), index=False) |
|
|
