A novel clinical scoring system for outcome prediction in dogs with acute kidney injury managed by hemodialysis

Gilad Segev, Philip H Kass, T. Francey, Larry D Cowgill

Research output: Contribution to journalArticle

44 Citations (Scopus)

Abstract

Background: No reliable tool to predict outcome of acute kidney injury (AKI) exists. Hypothesis: A statistically derived scoring system can accurately predict outcome in dogs with AKI managed with hemodialysis. Animals: One hundred and eighty-two client-owned dogs with AKI. Methods: Logistic regression analyses were performed initially on clinical variables available on the 1st day of hospitalization for relevance to outcome. Variables with P < .1 were considered for further analyses. Continuous variables outside the reference range were divided into quartiles to yield quartile-specific odds ratios (ORs) for survival. Models were developed by incorporating weighting factors assigned to each quartile based on the OR, using either the integer value of the OR (Model A) or the exact OR (Models B or C, when the etiology was known). A predictive score for each model was calculated for each dog by summing all weighting factors. In Model D, actual values for continuous variables were used in a logistic regression model. Receiver-operating curve analyses were performed to assess sensitivities, specificities, and optimal cutoff points for all models. Results: Higher scores were associated with decreased probability of survival (P < .001). Models A, B, C, and D correctly classified outcomes in 81, 83, 87, and 76% of cases, respectively, and optimal sensitivities/ specificities were 77/85, 81/85, 83/90 and 92/61%, respectively. Conclusions and Clinical Relevance: The models allowed outcome prediction that corresponded with actual outcome in our cohort. However, each model should be validated further in independent cohorts. The models may also be useful to assess AKI severity.

Original languageEnglish (US)
Pages (from-to)301-308
Number of pages8
JournalJournal of Veterinary Internal Medicine
Volume22
Issue number2
DOIs
StatePublished - Mar 2008

Fingerprint

hemodialysis
Acute Kidney Injury
Renal Dialysis
Odds Ratio
kidneys
Dogs
prediction
dogs
Logistic Models
odds ratio
Sensitivity and Specificity
Reference Values
Hospitalization
Regression Analysis
etiology

Keywords

  • Acute renal failure
  • Canine
  • Risk factors
  • Score
  • Survival

ASJC Scopus subject areas

  • veterinary(all)

Cite this

A novel clinical scoring system for outcome prediction in dogs with acute kidney injury managed by hemodialysis. / Segev, Gilad; Kass, Philip H; Francey, T.; Cowgill, Larry D.

In: Journal of Veterinary Internal Medicine, Vol. 22, No. 2, 03.2008, p. 301-308.

Research output: Contribution to journalArticle

@article{2b6c8d81c14447f295e7c78b6b195746,
title = "A novel clinical scoring system for outcome prediction in dogs with acute kidney injury managed by hemodialysis",
abstract = "Background: No reliable tool to predict outcome of acute kidney injury (AKI) exists. Hypothesis: A statistically derived scoring system can accurately predict outcome in dogs with AKI managed with hemodialysis. Animals: One hundred and eighty-two client-owned dogs with AKI. Methods: Logistic regression analyses were performed initially on clinical variables available on the 1st day of hospitalization for relevance to outcome. Variables with P < .1 were considered for further analyses. Continuous variables outside the reference range were divided into quartiles to yield quartile-specific odds ratios (ORs) for survival. Models were developed by incorporating weighting factors assigned to each quartile based on the OR, using either the integer value of the OR (Model A) or the exact OR (Models B or C, when the etiology was known). A predictive score for each model was calculated for each dog by summing all weighting factors. In Model D, actual values for continuous variables were used in a logistic regression model. Receiver-operating curve analyses were performed to assess sensitivities, specificities, and optimal cutoff points for all models. Results: Higher scores were associated with decreased probability of survival (P < .001). Models A, B, C, and D correctly classified outcomes in 81, 83, 87, and 76{\%} of cases, respectively, and optimal sensitivities/ specificities were 77/85, 81/85, 83/90 and 92/61{\%}, respectively. Conclusions and Clinical Relevance: The models allowed outcome prediction that corresponded with actual outcome in our cohort. However, each model should be validated further in independent cohorts. The models may also be useful to assess AKI severity.",
keywords = "Acute renal failure, Canine, Risk factors, Score, Survival",
author = "Gilad Segev and Kass, {Philip H} and T. Francey and Cowgill, {Larry D}",
year = "2008",
month = "3",
doi = "10.1111/j.1939-1676.2008.0063.x",
language = "English (US)",
volume = "22",
pages = "301--308",
journal = "Journal of Veterinary Internal Medicine",
issn = "0891-6640",
publisher = "Wiley-Blackwell",
number = "2",

}

TY - JOUR

T1 - A novel clinical scoring system for outcome prediction in dogs with acute kidney injury managed by hemodialysis

AU - Segev, Gilad

AU - Kass, Philip H

AU - Francey, T.

AU - Cowgill, Larry D

PY - 2008/3

Y1 - 2008/3

N2 - Background: No reliable tool to predict outcome of acute kidney injury (AKI) exists. Hypothesis: A statistically derived scoring system can accurately predict outcome in dogs with AKI managed with hemodialysis. Animals: One hundred and eighty-two client-owned dogs with AKI. Methods: Logistic regression analyses were performed initially on clinical variables available on the 1st day of hospitalization for relevance to outcome. Variables with P < .1 were considered for further analyses. Continuous variables outside the reference range were divided into quartiles to yield quartile-specific odds ratios (ORs) for survival. Models were developed by incorporating weighting factors assigned to each quartile based on the OR, using either the integer value of the OR (Model A) or the exact OR (Models B or C, when the etiology was known). A predictive score for each model was calculated for each dog by summing all weighting factors. In Model D, actual values for continuous variables were used in a logistic regression model. Receiver-operating curve analyses were performed to assess sensitivities, specificities, and optimal cutoff points for all models. Results: Higher scores were associated with decreased probability of survival (P < .001). Models A, B, C, and D correctly classified outcomes in 81, 83, 87, and 76% of cases, respectively, and optimal sensitivities/ specificities were 77/85, 81/85, 83/90 and 92/61%, respectively. Conclusions and Clinical Relevance: The models allowed outcome prediction that corresponded with actual outcome in our cohort. However, each model should be validated further in independent cohorts. The models may also be useful to assess AKI severity.

AB - Background: No reliable tool to predict outcome of acute kidney injury (AKI) exists. Hypothesis: A statistically derived scoring system can accurately predict outcome in dogs with AKI managed with hemodialysis. Animals: One hundred and eighty-two client-owned dogs with AKI. Methods: Logistic regression analyses were performed initially on clinical variables available on the 1st day of hospitalization for relevance to outcome. Variables with P < .1 were considered for further analyses. Continuous variables outside the reference range were divided into quartiles to yield quartile-specific odds ratios (ORs) for survival. Models were developed by incorporating weighting factors assigned to each quartile based on the OR, using either the integer value of the OR (Model A) or the exact OR (Models B or C, when the etiology was known). A predictive score for each model was calculated for each dog by summing all weighting factors. In Model D, actual values for continuous variables were used in a logistic regression model. Receiver-operating curve analyses were performed to assess sensitivities, specificities, and optimal cutoff points for all models. Results: Higher scores were associated with decreased probability of survival (P < .001). Models A, B, C, and D correctly classified outcomes in 81, 83, 87, and 76% of cases, respectively, and optimal sensitivities/ specificities were 77/85, 81/85, 83/90 and 92/61%, respectively. Conclusions and Clinical Relevance: The models allowed outcome prediction that corresponded with actual outcome in our cohort. However, each model should be validated further in independent cohorts. The models may also be useful to assess AKI severity.

KW - Acute renal failure

KW - Canine

KW - Risk factors

KW - Score

KW - Survival

UR - http://www.scopus.com/inward/record.url?scp=42449130592&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=42449130592&partnerID=8YFLogxK

U2 - 10.1111/j.1939-1676.2008.0063.x

DO - 10.1111/j.1939-1676.2008.0063.x

M3 - Article

VL - 22

SP - 301

EP - 308

JO - Journal of Veterinary Internal Medicine

JF - Journal of Veterinary Internal Medicine

SN - 0891-6640

IS - 2

ER -