Administrative databases contain precious information that can support the identification of specific pathologies. Specifically, chronic kidney disease (CKD) patients could be identified using hospital discharge records (HDR); these should contain information on the CKD stage using subcategories of the ICD9-CM classification's 585 code (subcategories can be expressed just by adding a fourth digit to this code). To verify the accuracy of HDR data regarding the coding of CKD collected in the Italian region Emilia-Romagna, we analyzed the HDR records of patients enrolled in the PIRP project, which could easily be matched with eGFR data obtained through laboratory examinations. The PIRP database was used as the gold standard because it contains data on CKD patients followed up since 2004 in thirteen regional nephrology units and includes data obtained from reliable and homogeneous laboratory measurement. All HDR of PIRP patients enrolled between 2009 and 2017 were retrieved and matched with available laboratory data on eGFR, collected within 15 days before or after discharge. We analyzed 4.168 HDR, which were classified as: a) unreported CKD (n=1.848, 44.3%); b) unspecified CKD, when code 585.9 (CKD, not specified) or 586 was used (n=446, 10.7%); c) wrong CKD (n=833, 20.0%); d) correct CKD (n=1041, 25.0%). We noticed the proportion of unreported CKD growing from 32.9% in 2009 to 56.6% in 2017, and the correspondent proportion of correct CKDs decreasing from 25.4% to 22.3%. Across disciplines, Nephrology showed the highest concordance (69.1%) between the CKD stage specified in the HDRs and the stage reported in the matched laboratory exam, while none of the other disciplines, except for Geriatrics, reached 20% concordance. When the CKD stage was incorrectly coded, it was generally underestimated; among HDRs with unreported or unspecified CKD at least half of the discharges were matched with lab exams reporting CKD in stage 4 or 5. We found that the quality of CKD stage coding in the HDR record database was very poor, and insufficient to identify CKD patients unknown to nephrologists. Moreover, the growing proportion of unreported CKD could have an adverse effect on patients' timely referral to a nephrologist, since general practitioners might remain unaware of their patients' illness. Actions aimed at improving the training of the operators in charge of HDRs compilation and, most of all, at allowing the exploitation of the informative potential of HDRs for epidemiological research are thus needed.

Abstract I flussi dei dati amministrativi sanitari regionali possono costituire una fonte preziosa per intercettare patologie di interesse. In particolare, dalle Schede di Dimissione Ospedaliera (SDO) si potrebbero intercettare i pazienti con Malattia Renale Cronica (MRC) mediante l’uso della sottocategoria (codice a 4 cifre) del codice 585 della classificazione ICD9-CM. Con il presente studio abbiamo fornito una valutazione sulla qualità della codifica della MRC nelle SDO della regione Emilia-Romagna, usando come gold standard il database del progetto PIRP, che raccoglie dati di circa 30.000 pazienti con diagnosi accertata di MRC in fase conservativa e in cui la definizione dello stadio di malattia è basata sui valori degli esami di laboratorio. Sono stati selezionati tutti i ricoveri ospedalieri dei pazienti in MRC partecipanti al progetto PIRP effettuati nel periodo 1.1.2009-31.12.2017. La valutazione di qualità delle SDO si è basata su di una classificazione delle schede in quattro categorie esaustive in base al match con esami PIRP eseguiti entro 15 gg. (±) dal ricovero: a) MRC assente: le SDO in cui non era stato riportato in nessuna posizione della diagnosi un codice ICD9-CM di malattia renale cronica 585.x (Malattia renale cronica) o 586 (Insufficienza renale, non specificata); b) MRC non specificata: quando la MRC era stata riportata con i codici 585.9 (Malattia renale cronica, non specificata) oppure 586; c) MRC errata: le SDO in cui la quarta cifra del codice ICD9-CM etichettava uno stadio di MRC diverso da quello dell’esame PIRP e d) MRC corretta: quelle in cui i due stadi corrispondevano. Nelle 4.168 SDO selezionate, 1.848 (44.3%) non includevano alcun codice di MRC; le SDO con MRC non specificata erano 446 (10.7%); con MRC errata erano 833 (20.0%) e con MRC corretta erano 1.041 (25.0%). Il trend temporale delle MRC corrette appare in decremento dopo un iniziale miglioramento (22.3% nel 2017), mentre al contrario è cresciuta dal 32.9% del 2009 al 56.6% del 2017 la quota di MRC assenti. Tra le discipline di dimissione, Nefrologia aveva la più alta percentuale di MRC corrette (69.1%), e solamente Geriatria superava il 20% tra le altre discipline con almeno 50 schede di dimissione. La bassa qualità della codifica SDO per la MRC tra i ricoveri eseguiti in reparti diversi dalla Nefrologia preclude sostanzialmente la possibilità di utilizzare le SDO come fonte per identificare pazienti con MRC non noti ai nefrologi. Inoltre, la MRC non compare nella lettera di dimissione dal ricovero se non viene inclusa nelle SDO, e così il medico di famiglia perde un’informazione che con ogni probabilità lo avrebbe motivato ad inviare il paziente presso gli specialisti nefrologi. Occorre pertanto intensificare la formazione degli addetti alla compilazione delle SDO, sensibilizzandoli all’utilizzo della quarta cifra, nonché “rivisitare” il significato delle SDO, che hanno molte potenzialità per la ricerca epidemiologica. La qualità della codifica SDO potrebbe inoltre migliorare sensibilmente se l’input del valore della creatinina venisse reso obbligatorio per tutte le discipline.

The accuracy of hospital discharge records and their use in identifying and staging chronic kidney disease

Gibertoni D.;Mandreoli M.;Caruso F.;Gasperoni L.;Orrico C.;Martelli D.;Ferri B.;Flachi M.;Iommi M.;
2019

Abstract

Administrative databases contain precious information that can support the identification of specific pathologies. Specifically, chronic kidney disease (CKD) patients could be identified using hospital discharge records (HDR); these should contain information on the CKD stage using subcategories of the ICD9-CM classification's 585 code (subcategories can be expressed just by adding a fourth digit to this code). To verify the accuracy of HDR data regarding the coding of CKD collected in the Italian region Emilia-Romagna, we analyzed the HDR records of patients enrolled in the PIRP project, which could easily be matched with eGFR data obtained through laboratory examinations. The PIRP database was used as the gold standard because it contains data on CKD patients followed up since 2004 in thirteen regional nephrology units and includes data obtained from reliable and homogeneous laboratory measurement. All HDR of PIRP patients enrolled between 2009 and 2017 were retrieved and matched with available laboratory data on eGFR, collected within 15 days before or after discharge. We analyzed 4.168 HDR, which were classified as: a) unreported CKD (n=1.848, 44.3%); b) unspecified CKD, when code 585.9 (CKD, not specified) or 586 was used (n=446, 10.7%); c) wrong CKD (n=833, 20.0%); d) correct CKD (n=1041, 25.0%). We noticed the proportion of unreported CKD growing from 32.9% in 2009 to 56.6% in 2017, and the correspondent proportion of correct CKDs decreasing from 25.4% to 22.3%. Across disciplines, Nephrology showed the highest concordance (69.1%) between the CKD stage specified in the HDRs and the stage reported in the matched laboratory exam, while none of the other disciplines, except for Geriatrics, reached 20% concordance. When the CKD stage was incorrectly coded, it was generally underestimated; among HDRs with unreported or unspecified CKD at least half of the discharges were matched with lab exams reporting CKD in stage 4 or 5. We found that the quality of CKD stage coding in the HDR record database was very poor, and insufficient to identify CKD patients unknown to nephrologists. Moreover, the growing proportion of unreported CKD could have an adverse effect on patients' timely referral to a nephrologist, since general practitioners might remain unaware of their patients' illness. Actions aimed at improving the training of the operators in charge of HDRs compilation and, most of all, at allowing the exploitation of the informative potential of HDRs for epidemiological research are thus needed.
2019
Gibertoni D.; Mandreoli M.; De Amicis S.; Cantarelli C.; Corradini M.; Caruso F.; Testa F.; Gasperoni L.; Orrico C.; Brancaleoni F.; Martelli D.; Angelini M.L.; Ferri B.; Flachi M.; Iommi M.; Santoro A.
File in questo prodotto:
Eventuali allegati, non sono esposti

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11585/725235
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo

Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 1
  • ???jsp.display-item.citation.isi??? ND
social impact