Trends in Hospital Admission forDiabetic Ketoacidosis in AdultsWith Type 1 andType 2Diabetes inEngland, 1998–2013: ARetrospective Cohort StudyDiabetes Care 2018;41:1870–1877 | https://doi.org/10.2337/dc17-1583

OBJECTIVE

This study determined trends in hospital admission for diabetic ketoacidosis (DKA) inadults with type 1 diabetes mellitus (T1DM) and type 2 diabetes mellitus (T2DM)from 1998 to 2013 in England.

RESEARCH DESIGN AND METHODS

The study population included 23,246 adults with T1DM and 241,441 adults withT2DM from the Clinical Practice Research Datalink (CPRD) and Hospital EpisodeStatistics (HES). All hospital admissions for DKA as the primary diagnosis from1998 to 2013 were identified. Trends in hospital admission for DKA in incidence,length of hospital stay, 30-day all-cause readmission rate, and 30-day and 1-year all-causemortality rates were determined using joinpoint regression, negative binomialregression, and logistic regression models.

RESULTS

For T1DM, the incidence of hospital admission for DKA increased between 1998 and2007 and remained static until 2013. The incidence in 2013 was higher than that in1998 (incidence rate ratio 1.53 [95% CI 1.09–2.16]). For T2DM, the incidence in-creased 4.24% (2.82–5.69) annually between 1998 and 2013. The length of hospitalstay decreased over time for both diabetes types (P £ 0.0004). Adults with T1DMwere more likely to be discharged within 2 days compared with adults with T2DM(odds ratio [OR] 1.28 [1.07–1.53]). The 30-day readmission rate was higher in T1DMthan in T2DM (OR 1.61 [1.04–2.50]) but remained unchanged for both diabetes typesover time. Trends in 30-day and 1-year all-cause mortality rates were also stable,with no difference by diabetes type.

CONCLUSIONS

In the previous two decades in England, hospitalization for DKA increased in adultswith T1DMand in those with T2DM, and associated health care performance did notimprove except decreased length of hospital stay.

Hospital admission for diabetic ketoacidosis (DKA) is a potentially life-threatening butlargely preventable complication in diabetes (1–3) and is associated with significanthealth care utilization and expenditure (4,5). Previous studies focused on describingDKA trends in children (6–9); however, .80% of patients with DKA are adults (1),

1Department of Nutrition, Gillings School ofGlobal Public Health, University of North Caro-lina, Chapel Hill, NC2Department of Preventive Medicine, North-western University Feinberg School of Medicine,Chicago, IL3Global Pharmacovigilance and Epidemiology,Sanofi US, Bridgewater, NJ4Department of Medicine, School of Medicine,University of North Carolina, Chapel Hill, NC

Corresponding author: Elizabeth J. Mayer-Davis,mayerdav@email.unc.edu.

Received 31 July 2017 and accepted 30 Decem-ber 2017.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc17-1583/-/DC1.

© 2018 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered. More infor-mation is available at http://www.diabetesjournals.org/content/license.

See accompanying article, p. 1839.

Victor W. Zhong,1,2 Juhaeri Juhaeri,3 and

Elizabeth J. Mayer-Davis1,4

1870 Diabetes Care Volume 41, September 2018

EPIDEM

IOLO

GY/HEA

LTHSERVICES

RESEA

RCH

among whom data for long-term trendsare limited. The available studies in adultsfrom different countries reported in-creased (10,11), stable (12), or slightly de-clined (13) rates of DKA admission to thehospital, emergency department, or in-tensive care unit. Other studies in adultsfound declined hospitalization rates forhyperglycemia as a whole (14,15). How-ever, none of these studies assessed DKAtrends by diabetes type. Type 2 diabetesmellitus (T2DM) is the predominant dia-betes type in adults, whereas DKA risk issubstantially higher in type 1 diabetesmellitus (T1DM) than in T2DM (1). Thus,direct applicationofDKA trends identifiedfrom all peoplewith diabetes to each spe-cific diabetes type is misleading. Usingprimary care data from the Clinical PracticeResearch Datalink (CPRD) and linked sec-ondary care data from theHospital EpisodeStatistics (HES), we aimed to determineDKA trends in incidence, length of hospitalstay, 30-day all-cause readmission rate, and30-day and 1-year all-causemortality rates,from 1998 to 2013, in adults with T1DM orT2DM in England, respectively.

RESEARCH DESIGN AND METHODS

Data SourceThe CPRD, one of the largest electronicmedical record databases in the world,collects anonymous longitudinal patientrecords from 684 general practices fromEngland, Scotland, Wales, and NorthernIreland as of January2015. Patientswithinthe CPRD are broadly representative ofthe population in the U.K. in age, sex, andethnicity (16). The CPRD contains rich pri-mary care information including, but notlimited to, basic demographics, prescribedmedications, clinical diagnoses, and labo-ratory test results. The CPRD is linked to asecondary care database via a trustedthird party, the HES, which houses pa-tient-level information on all hospitaladmissions to National Health Servicehospitals in England. The study popula-tion was from 398 of the 684 CPRD prac-tices that agreed to be linked to the HESbetween 1 April 1997 and 31March 2014.The linked CPRD-HES population is repre-sentative of the entire CPRD population(17). Hospital admissions for DKA wereidentified from the HES, and all otherdata were extracted from the CPRD. Thestudy protocol (No. 15_259RA) was ap-proved by the Independent ScientificAdvisory Committee for Medicines &Healthcare Products Regulatory Agency

in London, U.K. A minor modification wasmade to the approved protocol, which didnot need an amendment submission. Weincluded two additional variables, length ofhospital stay and mortality, that measurequality of care (18,19).

T1DM and T2DM DefinitionThe definition for T1DM and T2DM waspreviously published (20). Specifically,among individuals with one or morediabetes-relateddiagnosis code,wedefinedT1DM as having any one of the followingcriteria: 1) one or more T1DM codes andinsulin use only; 2) one or more T1DMcodes and insulin use only on the diagno-sis date and noninsulin glucose-loweringdrug (NIGLD), if any, was introduced6 months later; 3) two or more insulinprescriptions only. T2DM was defined ashaving anyone of the following criteria: 1)two or more T2DM codes and no T1DMcode, regardless of drug use; 2) one ormore T2DM codes and no T1DM codeand NIGLD use only; 3) one or moreT2DM codes and no T1DM code and be-ing treated with both NIGLD and insulin,but NIGLD prescribed no later than insu-lin; 4) two or more classes of NIGLD; 5)two or more prescriptions of a noninsulinnonmetformin glucose-lowering drugonly. NIGLDs includedmetformin, sulfony-lureas, glinide, thiazolidinediones, inhibi-tors of dipeptidyl peptidase 4, inhibitors ofsodium-glucose cotransporter 2, glucagon-like peptide 1 receptor agonists, and acar-bose. Patients were excluded if they hadnonresearch quality data determinedby the CPRD team based on the prespeci-fied data quality metrics or had a recordof secondary diabetes, maturity onset di-abetes of young, latent autoimmune di-abetes in adults, or malnutrition-relateddiabetes.

Patient Follow-up and DKA DiagnosisThe full-year HES data were available be-tween 1998 and 2013. The follow-upstarted at the latest of the followingdates: 1 January 1998, first diabetes visit,patient registration, up to standard date(16), or 18 years old. Follow-up ended atthe earliest of the following dates: 31 De-cember 2013, death, transfer out, or lastdata collection for the practice. DKA(E10.1, E11.1, E12.1, E13.1, and E14.1)listed as primary diagnosis for hospitaladmission during follow-up was identi-fied. We included all DKA admissions foranalyses. Because DKA could be the first

presentation of diabetes (1,21), we alsoincluded DKA admissions with hospitaladmission date within 28 days beforethe first diabetes visit date documentedin the CPRD.

Statistical AnalysisAll analyses were performed separatelyby diabetes type. The incidence rate ofhospital admission forDKAwas calculatedby dividing the total number of DKA ad-missions by the total accumulated person-years with diabetes within each yearbetween 1998 and 2013. We computedthe age standardized incidence rate usingthe entire study sample as the referencepopulation, according to these age catego-ries in T1DM (18–24, 25–34, 35–49, $50years) and T2DM (18–49, 50–64, $65years). For T1DM, age standardized inci-dence rates were also given by sex (male,female), duration of recorded diabeteshistory (as an alternative for diabetes du-ration; ,1, 1–9, 10–19, and $20 years),Charlson comorbidity score (1, 2, 3,and $4) (22), and region (North East,North West, Yorkshire and The Humber,East Midlands, West Midlands, East ofEngland, South West, South Central, Lon-don, and South East Coast). For T2DM,age standardized incidence rates weregiven by sex, duration of recorded diabe-tes history (,1, 1–4, 5–9, and$10 years),Charlson comorbidity score (1, 2, 3,and$4), current use of glucose-loweringdrugs (sulfonylureas with or withoutNIGLD, insulin only, insulin and NIGLD,and other), and region.

Trends in the incidence for hospital ad-mission for DKA were estimated usingjoinpoint regression models (23). Annualpercentage change and 95% CI werequantified. Length of hospital stay in dayswas calculated for adults with T1DM orT2DM and was then categorized intofour categories (0–2 days, 3–6 days, $1but ,2 weeks, and $2 weeks). Alsocalculated were the 30-day all-cause re-admission rate and 30-day and 1-year all-causemortality rates. Trends in the lengthof hospital stay were determined by neg-ative binomial regression models. Trendsin the four categories of hospital staywere estimated using ordinal logistic re-gressionmodels. Trends in the 30-day all-cause readmission rate and 30-day and1-year all-cause mortality rates were ex-amined using logistic regression models.All models included calendar year, age,sex, duration of recorded diabetes history,

care.diabetesjournals.org Zhong, Juhaeri, and Mayer-Davis 1871

Charlson comorbidity score, and currentuse of glucose-lowering drugs (for T2DMonly). Nonlinearity was assessed by add-ing a quadratic term for year. Differencesbetween T1DM and T2DM (referencegroup) were also compared. Incidencerate ratios (IRRs) or odds ratios (ORs)and 95% CIs were presented. To accountfor DKA as a recurrent event, we appliedrobust error estimate for correlated datain all models implemented by using thePROC GENMOD procedure with theREPEATED statement (24,25) in SAS 9.4software (SAS Institute). Statistical signif-icance was indicated by a two-sidedP value,0.05.Sensitivity analyses were conducted to

evaluate the effect of potential misclassi-fication of diabetes type on DKA trends.Previous investigation revealed that theaforementioned criteria 2 and 3 for T1DMand criterion 3 for T2DMwere mostly likelyto misclassify diabetes type from clinicalperspectives (20). A decent proportion ofadults with T1DM (11.86%) and a smallproportion of adults with T2DM (0.33%)may have been incorrectly classified. Wereestimated DKA trends by removingthesepatients.Wealso evaluatedwhetherDKA trends were materially influenced byany single year data.Our primary goal was to estimate the

burden of hospital admission for DKAover time in England, and thus, our pri-mary analyses focused on all availableDKA hospital admissions from each in-cluded patient. Secondary analyses wereconducted to determine whether the ob-servedDKA trendswere drivenby a groupofpatientswhohadmultipleDKAhospitaladmissions. We reassessed the trends by1) analyzing only the first DKA hospitaladmission from each patient during theentire study period and 2) analyzing onlythe first DKA hospital admission in eachyear from each patient. Characteristics ofpatients were also described according tothe number of DKA hospital admissionsduring the entire study period (none,one, two, three or more for T1DM andnone, one, and two or more for T2DM).Further, the length of hospital stay and30-day and 1-year all-cause mortalitywere compared between patients withdifferent numbers of DKA hospital admis-sions (one vs. two or three or more forT1DMand one vs. two ormore for T2DM),using the aforementioned modeling ap-proaches. In addition, among patientswith at least one DKA hospital admission,

we assessed the 30-day all-cause readmis-sion rate by including only the first 30-dayhospital readmission for DKA during theentire study period for each patient orthe first 30-day readmission in each yearfor each patient.

RESULTS

Among 23,246 adults with T1DM, 4,346hospital admissions for DKA were identi-fied during 121,258 person-years offollow-up (Supplementary Table 1). Theincidence of hospital admission for DKAwas 35.84 per 1,000 person-years duringthe entire study period. Among 241,441adults with T2DM, 1,148 DKA hospital ad-missions were found during 1,342,896

person-years of follow-up, with an inci-dence of 0.85 per 1,000 person-years(Supplementary Table 2). Every one infive DKA hospital admissions was attrib-uted to T2DM. The mean (SD) of age was46.66 (17.67) years in T1DM and 66.41(13.19) years in T2DM.

In adults with T1DM, the incidence forhospital admission for DKA did not haveany significant change between 1998 and2004 and then substantially increased by14.10% (95% CI 2.34–27.22, P = 0.02) an-nually from 2004 to 2007 (Fig. 1A). Theincidence then remained largely stable. Theincidence was much higher in 2013 thanin 1998 (IRR 1.53 [1.09–2.16]). Trends inDKA hospital admission increased during

Figure 1—Trends in hospital admission for DKA in adults with T1DM (A) or T2DM (B). Black dotsrepresent age standardized incidence rate in each year. Solid line represent estimated trend fromusing joinpoint regression. APC and 95% CI, along with P value for trend, are shown. The standardpopulation for calculating age standardized incidence was the entire cohort of adults with T1DM orT2DM during the study period. The following age categories were used: 18–24, 25–34, 35–49,and$50 years for T1DM, and 18–49, 50–64, and$65 years for T2DM. APC, annual percent change.

1872 Trends in Hospital Admission for DKA Diabetes Care Volume 41, September 2018

theentire studyperiod amongadultswithT1DM aged $35 years (SupplementaryFig. 1A and Supplementary Table 3),who were male (Supplementary Fig. 1B),with ,1 year or $20 years of recordeddiabetes history (Supplementary Fig. 1C),and few comorbidities (SupplementaryFig. 1D). A nonlinear trend was only seenin the South West region (SupplementaryFig. 1E). The incidence of hospital admis-sion for DKA was the highest amongadults with T1DM with ,1 year of re-corded diabetes history and young adultsaged 18–24 years.In adults with T2DM, the incidence of

hospital admission for DKA increasedfrom 0.70 to 0.98 per 1,000 person-years(Supplementary Table 2). The annual in-crease was 4.24% (95% CI 2.82–5.69, P,0.0001) (Fig. 1B). The incidence in-creased among those aged ,50 or $65(Supplementary Fig. 2A and Supplemen-tary Table 3), for both sexes (Supplemen-tary Fig. 2B), with$10 years of recordeddiabetes history (Supplementary Fig. 2C),and current insulin users (SupplementaryFig. 2E). The trend was stable amongsulfonylureas users. A decreasing trendwas observed among adults with T2DMwho were not currently being treatedwith insulin andsulfonylureas.Nononlineartrendwas foundby region (SupplementaryFig. 2F). Of note, the incidence of hospitaladmission for DKAwas the highest amonginsulin users and also high in young adultsand those with ,1 year or$10 years ofrecorded diabetes history.The median length of hospital stay in

days decreased over time for both T1DM(P , 0.0001) and T2DM (P = 0.0004)(Fig. 2A). No difference was found be-tween two diabetes types (rate ratio0.93 [95% CI 0.81–1.07]). When hospitalstay was examined as a categorical variable(Fig. 2B and C), adults with T1DM weremore likely to be dischargedwithin 2 days(OR 1.28 [1.07–1.53]) compared withadults with T2DM.The 30-day all-cause readmission rate

did not change over time for T1DM (P =0.94) and T2DM (P = 0.13), but the read-mission rate was higher in T1DM thanin T2DM (OR 1.61 [95% CI 1.04–2.50])(Fig. 3). Similarly, 30-day and 1-year all-cause mortality rates both remained un-changed for both diabetes types (P. 0.3)(Fig. 4A and B). The mortality rate did notdiffer by diabetes type.Sensitivity analyses revealed that after

excluding likely-to-be-misclassified adults

with T1DM, the incidence of hospital ad-mission for DKA increased 7.41% (95% CI5.23–9.65, P, 0.0001) annually between

1998 and 2009, and no significant changewas observed after then. After excludinglikely-to-be-misclassified adults with

Figure 2—Trends in median length of hospital stay (A) and by categories in adults with T1DM (B) orT2DM (C). Estimates were obtained from negative binomial models for A and from ordinal logisticregression models for B and C, with robust error variance to account for recurrent DKA admissions.Estimates were adjusted for calendar year, age, sex, duration of record diabetes history, Charlson comor-bidity score, and current use of glucose-lowering drugs (for T2DMonly). Adultswith T1DMweremore likelyto be discharged within 2 days (OR 1.28 [95% CI 1.07–1.53]) compared with adults with T2DM.

care.diabetesjournals.org Zhong, Juhaeri, and Mayer-Davis 1873

T2DM, the incidence increased 4.35%(2.80–5.92,P,0.0001) annually between1998 and 2013. For T1DM using the entiresample, if we excluded the year 2004data, an annual increase of 5.21% (2.67–6.77,P, 0.0001)was seen between1998and 2009, and then the incidence remainedunchanged (Supplementary Fig. 3).At least two hospital admissions for

DKA occurred in 3.17–4.42% of adultswith T1DM and 0.05–0.13% of adultswith T2DM in three selected years(Supplementary Table 4). Comparedwith those without DKA hospital admis-sion or with only one admission, adultswith T1DMwhohad recurrentDKAadmis-sions were younger (P , 0.0001), andadults with T2DM who had recurrentDKA admissions were younger, less likelyto be male, had longer duration of re-corded diabetes history, and were morelikely to be receiving insulin therapy (P,0.0001). The number of patients includedin each year and according to the numberof hospital admission for DKA were alsoshown for T1DM (Supplementary Table 5)and T2DM (Supplementary Table 6). Fur-ther analysis revealed that 65.72% and35.28% of the total DKA hospital admis-sions were recurrent for T1DM andT2DM, respectively.For T1DM, the incidence of hospital

admission for DKA in 2013 was 21.44,29.40, and 39.13 per 1,000 person-years,respectively, when including only the

first DKA admission during the entirestudy period (Supplementary Fig. 4A),the first DKA admission in each year(Supplementary Fig. 5A), and all DKA ad-missions (Fig. 1A) from each patient. TheDKA trends increased1.63% (95%CI0.32–2.95, P = 0.02) and 3.10% (1.93–4.28, P =0.0001) annually between1998and2013,respectively, when analyzing only the firstDKA admission during the entire studyperiod and the first DKA admission ineach year. For T2DM, the incidence ofhospital admission for DKA in 2013 was0.68, 0.84, and 0.98 per 1,000 person-years, respectively, when including onlythe first DKA admission during the entirestudy period (Supplementary Fig. 4B),the first DKA admission in each year(Supplementary Fig. 5B), and all DKA ad-missions (Fig. 1B) from each patient. TheDKAtrends increased2.02% (0.27–3.81, P =0.03) and 3.40% (1.82–5.00, P = 0.0004)annually between 1998 and 2013, re-spectively, when analyzing only the firstDKA admission during the entire studyperiod and the first DKA admission ineach year.

During the entire study period, the av-erage 30-day all-cause readmission ratedecreased from 12.67 to 6.05% forT1DM and from 6.18 to 3.74% for T2DMwhen comparing the analysis with includ-ing all DKA hospital admissions to includ-ing only the first DKA readmission foreach patient (Supplementary Table 7).

Similar to the primary analysis with allDKA admissions included (Fig. 3), thetrends in the 30-day all-cause readmis-sion rate remained stable over timefor both T1DM and T2DM (P . 0.1)(Supplementary Fig. 6A and B). Adultswith T1DM with three or more DKA ad-missions (OR1.90 [95%CI 1.15–3.13]) andadults with T2DM with two or more DKAadmissions (2.05 [1.16–3.62]) were asso-ciated with higher 1-year all-cause mor-tality compared with those adults withT1DM or T2DM with only one DKA ad-mission (Supplementary Table 8). No dif-ference was found for 30-day all-causemortality and length of hospital stay forboth T1DM and T2DM.

CONCLUSIONS

Among a large representative cohort ofadults with T1DM or T2DM in England,from 1998 to 2013, we found an in-creased incidence of hospital admissionfor DKA before 2007 and no change wasseen after then for adults with T1DM. Theincidence of hospital admission for DKAincreased steadily among adults withT2DM during the entire period. About20% of all DKA admissions were fromadults with T2DM. The length of hospitalstay for adults with T1DM or T2DM de-creased over time, but the 30-day all-cause readmission rate and 30-day and1-year all-cause mortality rates remainedunchanged. Nonetheless, adults withT1DM were more likely to be dischargedwithin 2 days and also had higher read-mission rate than adults with T2DM.These findings provide informative datato understand the long-term DKA burdenand associated health care performancein England and to inform future practice-level or policy-level intervention for re-ducing DKA burden.

DKA is a well-recognized hallmark ofT1DM, but our findings suggest thatman-aging DKA in adults with T1DM has notbeen successful in England. We foundthat the incidence of hospital admissionfor DKA was 53% higher in 2013 than in1998, although the incidence stopped in-creasing after 2007. Given the substan-tially higher DKA risk in T1DM than inT2DM, an elevated DKA trend in T1DMwould contribute considerable healthcare and patient burden. Importantly, theincreasing trend was the most prominentamong adults with T1DMwith,1 year ofrecorded diabetes history who also hadthe highest risk for DKA hospitalization.

Figure 3—Trends in 30-day all-cause readmission rate for hospital admission for DKA. Estimateswere obtained from logistic regression models with robust error variance to account for re-current DKA admissions. Estimates were adjusted for calendar year, age, sex, duration ofrecorded diabetes history, Charlson comorbidity score, and current use of glucose-loweringdrugs (for T2DM only).

1874 Trends in Hospital Admission for DKA Diabetes Care Volume 41, September 2018

This subgroup of patients requires partic-ular attention. For T2DM, most of theDKA admissions were from insulin users,and the increase in DKA incidence amonginsulin users was also the greatest amongall subgroups of adults with T2DM, sug-gesting that appropriate use of and ad-herence to insulin therapy have beenchallenging.Our data suggest that the incidence of

hospital admission for DKA was.40-foldhigher in T1DM than in T2DM, but everyone in fiveDKA admissionswas attributedto T2DM. These concur with the evidencefrom Wales, U.K. (11), the U.S. (26), andNordic countries (27,28). The contribu-tion of DKA burden from T2DM is moresignificant than traditionally recognized

when DKA was considered a key clini-cal feature for T1DM (2). In fact, werevealed a steady increase in the inci-dence of hospital admission for DKA inT2DM. The underlying causes are unclear,but rising T2DMprevalence and incidencemay be one (29). The risk of hospital ad-mission for DKA was higher among thosewith ,1 year or $10 years of recordeddiabetes history. In our study population,the proportion of adults with T2DMwith,1 year of recorded diabetes historydecreased over time (SupplementaryFig. 7), but more adults had $10 yearsof recorded diabetes history. Therefore,the aging of population with T2DM mayplay amore important role. Although spec-ulative, other reasons may be increased

insulin use and increased prevalence ofinfection and ketosis-prone T2DM overtime. Infection and noncompliance to in-sulin therapy are two of the most com-mon precipitating factors for DKA (3).Whether there is an increased trend ofinfection in T2DM patients paralleling in-creased DKA trends needs future investi-gation. Our data indicate that insulin usershad the greatest risk for DKA hospital ad-mission in T2DM. A previous study re-ported that the prevalence of insulin useincreased in earlier years before 2005 andthen remained in that high level (20). DKAis commonly seen in ketosis-prone T2DM,which usually occurs in nonwhite patients(2), but correctly phenotyping this sub-type is difficult.

Length of hospital stay and 30-day all-cause readmission andmortality rates arecommonly used outcome measures forhealth care utilization and performance(18,19). The declining trend in DKA-re-lated length of hospital stay found inour study may imply reduced hospitalcosts per DKA admission and improvedtreatment andmanagement. Discrepancyexisted by diabetes type. Adults withT1DM were much more likely to be dis-charged within 2 days (65% vs. 29% in2013), which may be partly related toadults with T2DM being;20 years olderthan adults with T1DM. Although specu-lative, the shortened hospital stay maypartly account for the unchanged 30-dayall-cause readmission rate and 30-dayand 1-year all-cause mortality rates, be-cause it may be likely that patients weredischarged too quickly before they werefully treated to reduce health carecosts. However, the temporal relation-ship between length of hospital stay,short-term readmission and mortalityrates, and hospital costs requires moreinvestigations.

Approximately two-thirds of all DKAhospital admissions in T1DM and one-third in T2DM (mostly insulin users)were recurrent cases in our study. There-fore, restricting analyses to include onlythe first DKA hospital admission and thefirst DKA readmission considerably re-duced the incidence rate and 30-day re-admission rate for DKA, respectively.Nonetheless, we still observed an eleva-tion of trends in hospital admission forDKA for both T1DM and T2DM duringthe entire study period after applyingthese restrictions. Other than previouslyhypothesized reasons for the rising DKA

Figure 4—Trends in the 30-day (A) and 1-year (B) all-cause mortality rate. Estimates were obtainedfrom logistic regressionmodels with robust error variance to account for recurrent DKA admissions.Estimates were adjusted for calendar year, age, sex, duration of recorded diabetes history, Charlsoncomorbidity score, and current use of glucose-lowering drugs (for T2DM only).

care.diabetesjournals.org Zhong, Juhaeri, and Mayer-Davis 1875

trends, the diagnostic criteria ofDKA fromthe AmericanDiabetes Association in 2001recommended a classificationof the sever-ity of DKA (mild, moderate, and severe) tohelp with patient disposition and selectionof therapy (30). The specific inclusion ofDKA with mild metabolic acidosis mayhave resulted in an increased number ofDKA cases, although patients with mildDKA can be treated and observed in theemergency department and may not beadmitted for hospitalization.Reducing DKA risk in adults has been

challenging. Possibly, managing chroniccomplications, such as retinopathy, ne-phropathy, neuropathy, and amputationand cardiovascular disease risk, are thepriorities, and cardiovascular disease isthe primary cause of death and source ofhealth care costs in people with T1DM orT2DM (31–33). However, managing acutecomplications such as DKA deservesmoreattention. Our study andmost of the pub-lished studies across different countriesreported increased/stable DKA rates oran increased number of hospital admis-sions, or both (10–12). Idealmanagementfor DKA should prevent both the first andrecurrent DKA and improve care duringhospitalization and immediately after dis-charge. Unlike hypoglycemia, which is to-tally preventable (34), some DKA casesmay not have an obvious precipitatingcause, particularly in ketosis-prone pa-tients with T2DM (1–3). Nonetheless,most DKA cases are preventable throughimplementing structured educationalprograms emphasizing the importanceof continuation of and adherence to in-sulin therapy, improving sick day man-agement (e.g., increase insulin doses ifunwell, test ketones), having medica-tions available for treating fever or in-fection, accessing medical care easily,and enhancing communication with ahealth care provider during an intercur-rent illness (1–3).Our study had several important limi-

tations. First, we did not have data onDKA events admitted to the emergencydepartment but not resulting in hospital-ization. Second, misclassification of diabe-tes type was likely. Sensitivity analysessuggested that the potential misclassifica-tion would not change our conclusions.However, we could not rule out the possi-bility of misdiagnosing ketosis-proneT2DM as T1DM, and thus, its effect onthe DKA trends in T1DM and T2DM wasunknown (2,35). Third, the identification

of DKA admissions was solely based onthe ICD-10 diagnosis code and was notverified by laboratory criteria such as glu-cose level, pH level, serum bicarbonatelevel, anion gap, and serum/urine ke-tones level. Accordingly, DKA could havebeen miscoded or misdiagnosed. Fourth,the introduction of the Quality and Out-comes Framework (QOF) in 2004 mayimprove data completeness and facilitateidentification of patients with diabetes inthe CPRD (36). Coincidentally, the great-est increase was observed between2004 and 2007 for adults with T1DM, al-though the incidence of hospital admis-sion for DKA already went up before2003. Our sensitivity analyses indicatedthat the trend before 2007 was largelyinfluenced by the year 2004 data; drop-ping this year yielded a significant in-crease between 1998 and 2007. Also,we did not observe any unusual trendsin 2004–2007 in the following variablesin T1DM adults, including the number ofincluded adults with T1DM, the numberof DKA admissions, person-years withdiabetes, percentage of adults withT1DMwith durationof recordeddiabeteshistory ,1 year, percentage male, andpercentage of patients aged 18–24(Supplementary Table 9). Taken together,it is unlikely that our conclusion regard-ing the trends in hospital admission forDKA in T1DM adults was biased. Fifth,extending current DKA trends beyond2013 should be done cautiously. Sev-eral changes were made in December2015 and May 2017 to update the 2009diabetes guideline from the National In-stitute for Health and Care Excellence inthe U.K. (37,38). For example, sodium-glucose cotransporter 2 inhibitors havebeen added to the glucose-lowering ther-apy algorithm, and the recommendedHbA1c target level is 6.5% (48 mmol/mol)and 7.0% (53 mmol/mol) if more thanone glucose-lowering drug is prescribed.It remains unknown whether thesechanges have influenced the DKA trendsbeyond 2013, although we do not antic-ipate any dramatic alteration in the DKAtrends between 2014 and the time of thepublication of this study. Sixth, DKAtrends from our study may not be appli-cable to health care systems beyond theU.K. Seventh, we did not correct formul-tiple comparisons, and thus, we could notrule out that some findings may be dueto chance. Lastly, we underestimated du-ration of diabetes, because we used the

first diabetes visit date as an approximatefor the diabetes diagnosis date. To avoidconfusion, we used duration of recordeddiabetes history throughout instead ofdiabetes duration.

In summary, reducing DKA hospital ad-mission is urgent for adults with T1DM orT2DM in England, given the high incidenceof hospital admission for DKA in T1DMand the increasing trend in T2DM. Adultswith T2DM contribute 20% of DKA admis-sions, although the admission rate is40-fold higher in adults with T1DM. Inaddition to preventing DKA hospital admis-sion, successful management also needsto reduce short-term all-cause readmis-sion and mortality rates for which stabletrends have been seen in England. Futureinvestigations are required to identify fac-tors that contribute to the increased DKArisk and unchanged readmission and mor-tality rates for DKA.

Acknowledgments. The authors sincerelythankMeera Kumar and Ling Zhang, from SanofiUS (Bridgewater, NJ), for extracting the CPRDand HES data.Funding.This studywas supportedby the SanofiGlobal Nutrition Scholars program at the Uni-versity of North Carolina at Chapel Hill.Duality of Interest. V.W.Z. received financialsupport from the Sanofi Global Nutrition Scholarsprogram, and J.J. is an employee of Sanofi US,Bridgewater, NJ. No other potential conflicts of in-terest relevant to this article were reported.Author Contributions. V.W.Z. analyzed thedata and wrote and revised the manuscript.V.W.Z. and J.J. acquired the data. V.W.Z. andE.J.M.-D. designed the study. All authors re-viewed and edited the manuscript and contrib-uted to discussion. V.W.Z. and E.J.M.-D. are theguarantorsof thisworkand,assuch,hadfullaccessto all the data in the study and take responsibilityfor the integrityof thedataand theaccuracyof thedata analysis.

References1. Kitabchi AE, Umpierrez GE,Miles JM, Fisher JN.Hyperglycemic crises in adult patients with diabe-tes. Diabetes Care 2009;32:1335–13432. Misra S, Oliver N, Dornhorst A. Diabetic ketoa-cidosis: not always due to type 1 diabetes. BMJ2013;346:f35013. Misra S, Oliver NS. Diabetic ketoacidosis inadults. BMJ 2015;351:h56604. HexN, Bartlett C,Wright D, TaylorM, Varley D.Estimating the current and future costs oftype 1 and type 2 diabetes in the UK, includingdirect health costs and indirect societal andproductivity costs. Diabet Med 2012;29:855–8625. Dhatariya KK, Skedgel C, Fordham R. The costof treating diabetic ketoacidosis in the UK: a na-tional surveyof hospital resource use. DiabetMed2017;34:1361–1366

1876 Trends in Hospital Admission for DKA Diabetes Care Volume 41, September 2018

6. Dabelea D, Rewers A, Stafford JM, et al.;SEARCH for Diabetes in Youth Study Group.Trends in the prevalence of ketoacidosis at diabe-tes diagnosis: the SEARCH for Diabetes in YouthStudy. Pediatrics 2014;133:e938–e9457. Hekkala A, Knip M, Veijola R. Ketoacidosis atdiagnosis of type 1 diabetes in children in north-ern Finland: temporal changes over 20 years. Di-abetes Care 2007;30:861–8668. Fritsch M, Schober E, Rami-Merhar B, Hofer S,Frohlich-Reiterer E, Waldhoer T; Austrian Diabe-tes Incidence Study Group. Diabetic ketoacidosisat diagnosis in Austrian children: a population-based analysis, 1989-2011. J Pediatr 2013;163:1484–1488.e19. Neu A, Hofer SE, Karges B, Oeverink R,Rosenbauer J, Holl RW; DPV Initiative and theGermanBMBF CompetencyNetwork for DiabetesMellitus. Ketoacidosis at diabetes onset is still fre-quent in children and adolescents: a multicenteranalysis of 14,664 patients from 106 institutions.Diabetes Care 2009;32:1647–164810. KoSH, LeeWY, Lee JH, et al. Clinical character-istics of diabetic ketoacidosis in Korea over thepast two decades. Diabet Med 2005;22:466–46911. Abdulrahman GO, Amphlett B, Okosieme OE.Trends in hospital admissions with diabetic ketoa-cidosis in Wales, 1999-2010. Diabetes Res ClinPract 2013;100:e7–e1012. Ginde AA, Pelletier AJ, Camargo CA Jr. Na-tional study of U.S. emergency department visitswith diabetic ketoacidosis, 1993-2003. DiabetesCare 2006;29:2117–211913. Liu CC, Chen KR, Chen HF, Huang HL, Ko MC,Li CY. Trends in hospitalization for diabetic ketoa-cidosis in diabetic patients in Taiwan: analysis ofnational claims data, 1997-2005. J Formos MedAssoc 2010;109:725–73414. Lombardo F, Maggini M, Gruden G, Bruno G.Temporal trend in hospitalizations for acute dia-betic complications: a nationwide study, Italy,2001-2010. PLoS One 2013;8:e6367515. Lipska KJ, Ross JS, Wang Y, et al. Nationaltrends in US hospital admissions for hyperglyce-mia and hypoglycemia among Medicare benefi-ciaries, 1999 to 2011. JAMA InternMed2014;174:1116–112416. Herrett E, Gallagher AM, Bhaskaran K, et al.Data resource profile: Clinical Practice ResearchDatalink (CPRD). Int J Epidemiol 2015;44:827–83617. Gallagher AM. Puri S, van Staa TP. Linkage ofthe General Practice Research Database (GPRD)

with other data sources. PharmacoepidemiolDrug Saf 2011;20:S230–S23118. Krumholz HM, Keenan PS, Brush JE Jr, et al.;American Heart Association InterdisciplinaryCouncil on Quality of Care and Outcomes Re-search; American College of Cardiology Founda-tion. Standards for measures used for publicreporting of efficiency in health care: a scientificstatement from the American Heart AssociationInterdisciplinary Council on Quality of Care andOutcomes Research and the American College ofCardiology Foundation. J Am Coll Cardiol 2008;52:1518–152619. Bueno H, Ross JS, Wang Y, et al. Trends inlength of stay and short-term outcomes amongMedicare patients hospitalized for heart failure,1993-2006. JAMA 2010;303:2141–214720. Zhong VW, Juhaeri J, Cole SR, et al. Incidenceand trends in hypoglycemia hospitalization inadults with type 1 and type 2 diabetes in England,1998-2013: a retrospective cohort study. DiabetesCare 2017;40:1651–166021. Hardern RD, Quinn ND. Emergency manage-ment of diabetic ketoacidosis in adults. EmergMed J 2003;20:210–21322. Khan NF, Perera R, Harper S, Rose PW. Adap-tation and validation of the Charlson Index forRead/OXMIS coded databases. BMC Fam Pract2010;11:123. National Cancer Institute, Division of CancerControl and Population Sciences. JoinpointTrend Analysis Software. Regression program,version 4.5.0.1 - June 2017. Available fromhttps://surveillance.cancer.gov/joinpoint/. Accessed7 November 201724. SAS Institute. Generalized estimating equa-tions. SAS/STAT(R) 9.2 user’s guide, second edi-tion. Available from https://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_genmod_sect001.htm. Accessed 7 November 201725. Zou G. A modified Poisson regression ap-proach to prospective studies with binary data.Am J Epidemiol 2004;159:702–70626. Newton CA, Raskin P. Diabetic ketoacidosis intype 1 and type 2 diabetes mellitus: clinical andbiochemical differences. Arch Intern Med 2004;164:1925–193127. Wang ZH, Kihl-Selstam E, Eriksson JW. Ketoa-cidosis occurs in both type 1 and type 2 diabe-tes–a population-based study from NorthernSweden. Diabet Med 2008;25:867–870

28. Henriksen OM, RøderME, Prahl JB, SvendsenOL. Diabetic ketoacidosis in Denmark: incidenceand mortality estimated from public health regis-tries. Diabetes Res Clin Pract 2007;76:51–5629. Sharma M, Nazareth I, Petersen I. Trends inincidence, prevalence and prescribing in type 2diabetes mellitus between 2000 and 2013 in pri-mary care: a retrospective cohort study [pub-lished correction appears in BMJ Open 2016;6:e010210corr1]. BMJ Open 2016;6:e01021030. Kitabchi AE, Umpierrez GE,MurphyMB, et al.Management of hyperglycemic crises in patientswith diabetes. Diabetes Care 2001;24:131–15331. Laing SP, Swerdlow AJ, Slater SD, et al. Mor-tality from heart disease in a cohort of 23,000patients with insulin-treated diabetes. Diabetolo-gia 2003;46:760–76532. Paterson AD, Rutledge BN, Cleary PA, LachinJM, Crow RS; Diabetes Control and ComplicationsTrial/Epidemiology of Diabetes Interventions andComplications Research Group. The effect of in-tensive diabetes treatment on resting heart ratein type 1 diabetes: the Diabetes Control and Com-plications Trial/Epidemiology of Diabetes Inter-ventions and Complications study. Diabetes Care2007;30:2107–211233. Hogan P, Dall T, Nikolov P; AmericanDiabetesAssociation. Economic costs of diabetes in the USin 2002. Diabetes Care 2003;26:917–93234. Seaquist ER, Anderson J, Childs B, et al. Hypo-glycemia and diabetes: a report of a workgroup ofthe American Diabetes Association and the Endo-crine Society. Diabetes Care 2013;36:1384–139535. Xie XJ, Hu Y, Cheng C, Feng TT, He K,Mao XM.Should diabetic ketosis without acidosis be in-cluded in ketosis-prone type 2 diabetes mellitus?Diabetes Metab Res Rev 2014;30:54–5936. Kontopantelis E, Reeves D, Valderas JM,Campbell S, Doran T. Recorded quality of primarycare for patients with diabetes in England beforeand after the introduction of a financial incentivescheme: a longitudinal observational study. BMJQual Saf 2013;22:53–6437. National Institute for Health and Care Excel-lence. Type 2 diabetes in adults: management.NG28.December2015 (updatedMay2017).Avail-able from https://www.nice.org.uk/guidance/ng28. Accessed 24 December 201738. McGuire H, Longson D, Adler A, Farmer A,Lewin I; Guideline Development Group. Manage-ment of type 2 diabetes in adults: summary ofupdated NICE guidance. BMJ 2016;353:i1575

care.diabetesjournals.org Zhong, Juhaeri, and Mayer-Davis 1877