Ben K. Graf, MD

Observation hospitalization is a growing phenomenon in the United States healthcare system.[1] For Medicare beneficiaries, observation encounters increased 33.6% from 2004 to 2011, with inpatient encounters decreasing by 7.8% over the same period.[2]

Observation length of stay has also increased. Medicare states that observation care should typically last <24 hours, and in only rare and exceptional cases exceed 48 hours. We and others have showed that observation stays, on average, do not meet this definition.[1, 3] At our institution, historically less than half of observation encounters discharge in under 24 hours, and 1 in 6 stay longer than 48 hours.[3]

The Centers for Medicare and Medicaid Services (CMS) issued a rules change effective October 1, 2013, in response to concern about recent increases in the length of time that CMS beneficiaries spend as hospital outpatients receiving observation services.[4] These rules shifted observation determination from clinical criteria, such as InterQual,[5] to a time‐based rule that hinges on a 2‐midnight cut point. Patients staying <2 midnights, with few exceptions, are now observation, and those staying 2 midnights are inpatients. This is important, as patients hospitalized as observation are technically outpatients, not covered by Medicare Part A hospital insurance, and ineligible for skilled nursing facility benefits.[3, 6]

Although challenges with observation status in general are well described,[1, 3, 7] the potential impact of the 2‐midnight rule is not fully known. The purpose of our descriptive study was to examine how the new rules, retrospectively applied to recent encounters, would affect our hospital encounters, with a separate analysis of Medicare encounters and adult general medicine encounters. First, as CMS predicted a net shift from outpatient to inpatient stays[8] under the new rule, we attempted to determine whether this rule would actually reduce observation encounter frequency. Second, as CMS has cited persistently large improper payment rates in short‐stay hospital inpatient claims[4] and intends to audit <2‐midnight inpatient encounters under the assumption that many are misclassified observation stays,[4, 8, 9] we sought to determine if short‐stay inpatient and observation encounters were truly the same. Third, as insurance coverage will change based on the 2‐midnight cut point, we sought to determine whether 2 midnights separated distinct clinical populations within observation status, making the rule logical and fair. Finally, we sought to determine whether external factors, such as time of admission, day of week of admission, and transfer to our institution would impact how patients may be classified under the new rule.

METHODS

RESULTS

Percentage of Observation Encounters Reaching 2 Midnights Based on Time of Hospitalization

Observation encounters starting before 8:00 _am spanned 2 midnights 13.6% of the time, and those encounters starting after 4:00 _pm crossed 2 midnights 31.2% of the time. Two of the 3 top ICD‐9 codes were the same across LOS, with similar findings for the adult general medicine‐only group (Figure 1).

Figure 1

DISCUSSION

Although CMS predicts that more patients will be classified as inpatients under the new rule, we determined the opposite, consistent with a recent report generated by the Office of the Inspector General (OIG) for 2012 Medicare beneficiaries.[8] Our results did not change when we excluded all surgical encounters to account for possible exclusions based on the surgical inpatient‐only list. Although a small percentage of Medicare adult general medicine patients may be reclassified as inpatients under the new rules, the net effect would be that many more hospital encounters will be billed under observation rather than inpatient status. These findings assume overall length of stay will remain unchanged under these rules, an assumption that may not hold true given the financial losses we predicted hospitals may face under this rule,[12] and potential pressures on individual physicians providing patient care.

Medicare has prioritized auditing <2‐midnight inpatient encounters under the assumption that many short inpatient encounters are actually misclassified observation encounters,[4, 8, 9] prompting us to investigate whether this was the case in our patient population. Although it did not use ICD‐9 diagnosis codes, the OIG report suggested that short‐stay inpatients and observation patients may be clinically similar.[8, 13] Using ICD‐9 codes, we found no overlap between the top ICD‐9 codes for adult general medicine <2‐midnight inpatient and observation encounters, and only 1 of 5 shared codes for these encounters across all service lines. These findings are counter to the OIG report, and suggest that <2‐midnight inpatients are different from observation patients at our institution, and that <2‐midnight inpatients should not be arbitrarily reclassified as observation based solely upon LOS.

We also found that the majority of top ICD‐9 codes within observation were the same regardless of LOS, suggesting that LOS does not reliably differentiate clinically different observation populations that merit different insurance coverage (Medicare Part A for 2‐midnight encounters, Medicare Part B for <2‐midnight encounters). This lack of a clear cut point may drive an overall increase in LOS to achieve 2 midnights, as common diagnosis codes can be justified for both <2‐midnight or 2‐midnight observation stays.

Finally, we found that external factors, such as the time of day and specific day (weekday vs weekend) of hospitalization, impact the likelihood of achieving a 2‐midnight stay. Patients hospitalized earlier in the day were less likely to span 2 midnights compared to later‐day encounters, suggesting that use of a full working day as the day of presentation is harmful to a patient's chance of gaining inpatient status. Observation hospitalizations starting on a weekend day were more likely to achieve 2 midnights, which likely reflects different resource allocation and hospital efficiency on weekends, yet it is unlikely that weekend midnights, if associated with any delay in care, will be counted toward a patient's cumulative 2‐midnight total. The CMS has further indicated that midnights accrued prior to transfer from 1 hospital to another will not count toward a cumulative 2‐midnight stay. Although it would seem likely that patients requiring transfer to a tertiary care center would meet inpatient criteria, 9% of our observation encounters were transferred from other acute care hospitals, and many will have lost attributable midnights accrued prior to transfer to our hospital. Taken together, our findings strongly suggest that issues entirely unrelated to diagnosis or clinical status will impact whether hospitalized patients will be classified as inpatient or observation under the new rule.

This study has several limitations. Our data were limited to a single Midwestern tertiary care academic medical center, and may not be applicable to other healthcare settings. Second, 1.5% (466) of our inpatient stays and 56 (1.2%) of our observation encounters lacked a referral source in our administrative database, although these deficiencies would not likely change the conclusions. Finally, the total number of observation encounters starting before 8:00 _am and staying 2 midnights was small and therefore potentially subject to confounding. However, despite the fact that encounters beginning in the early morning hours may be different in ways unable to be measured in this study, encounter ICD‐9 codes were similar regardless of time of day.

Despite these limitations, our study raises concerns about the impact of CMS's new time‐driven observation rules on hospital practice patterns. We found distinctly different <2‐midnight inpatient ICD‐9 codes as compared to those for observation encounters. Reclassifying <2‐midnight inpatients as observation may inappropriately shift the financial burden of acute hospitalizations to patients who were previously eligible to receive Medicare inpatient benefits. We also demonstrated a decrease in billable inpatient encounters under the new rules, which may be countered by changes in hospital and provider practice to drive more admissions over the 2‐midnight threshold to avoid financial losses. These changes should be defensible under audit given the ambiguities of the rule we found in this study surrounding time of day of admission, weekend and transfer hospitalizations, and the fact that common observation ICD‐9 codes are similar across LOS. Ironically, the unintended consequence of the new rule may be to drive up hospital LOS, reduce efficiency, and increase the overall cost of care.

Our findings suggest that CMS should define observation care per its original intent: as a means to determine if a patient can safely return home after a short period of additional care.[6] Patients whose conditions necessitate an intensity and level of service beyond this narrow scope should be classified as full inpatients irrespective of LOS, whether that LOS is <2 midnights or 2 midnights. Policies that fail to reflect the original intent of observation status are unlikely to achieve the intended outcome of defining 2 distinct patient populations that merit different services and reimbursement.

Observation hospitalization is a growing phenomenon in the United States healthcare system.[1] For Medicare beneficiaries, observation encounters increased 33.6% from 2004 to 2011, with inpatient encounters decreasing by 7.8% over the same period.[2]

Observation length of stay has also increased. Medicare states that observation care should typically last <24 hours, and in only rare and exceptional cases exceed 48 hours. We and others have showed that observation stays, on average, do not meet this definition.[1, 3] At our institution, historically less than half of observation encounters discharge in under 24 hours, and 1 in 6 stay longer than 48 hours.[3]

The Centers for Medicare and Medicaid Services (CMS) issued a rules change effective October 1, 2013, in response to concern about recent increases in the length of time that CMS beneficiaries spend as hospital outpatients receiving observation services.[4] These rules shifted observation determination from clinical criteria, such as InterQual,[5] to a time‐based rule that hinges on a 2‐midnight cut point. Patients staying <2 midnights, with few exceptions, are now observation, and those staying 2 midnights are inpatients. This is important, as patients hospitalized as observation are technically outpatients, not covered by Medicare Part A hospital insurance, and ineligible for skilled nursing facility benefits.[3, 6]

Although challenges with observation status in general are well described,[1, 3, 7] the potential impact of the 2‐midnight rule is not fully known. The purpose of our descriptive study was to examine how the new rules, retrospectively applied to recent encounters, would affect our hospital encounters, with a separate analysis of Medicare encounters and adult general medicine encounters. First, as CMS predicted a net shift from outpatient to inpatient stays[8] under the new rule, we attempted to determine whether this rule would actually reduce observation encounter frequency. Second, as CMS has cited persistently large improper payment rates in short‐stay hospital inpatient claims[4] and intends to audit <2‐midnight inpatient encounters under the assumption that many are misclassified observation stays,[4, 8, 9] we sought to determine if short‐stay inpatient and observation encounters were truly the same. Third, as insurance coverage will change based on the 2‐midnight cut point, we sought to determine whether 2 midnights separated distinct clinical populations within observation status, making the rule logical and fair. Finally, we sought to determine whether external factors, such as time of admission, day of week of admission, and transfer to our institution would impact how patients may be classified under the new rule.

METHODS

RESULTS

Percentage of Observation Encounters Reaching 2 Midnights Based on Time of Hospitalization

Observation encounters starting before 8:00 _am spanned 2 midnights 13.6% of the time, and those encounters starting after 4:00 _pm crossed 2 midnights 31.2% of the time. Two of the 3 top ICD‐9 codes were the same across LOS, with similar findings for the adult general medicine‐only group (Figure 1).

Figure 1

DISCUSSION

Although CMS predicts that more patients will be classified as inpatients under the new rule, we determined the opposite, consistent with a recent report generated by the Office of the Inspector General (OIG) for 2012 Medicare beneficiaries.[8] Our results did not change when we excluded all surgical encounters to account for possible exclusions based on the surgical inpatient‐only list. Although a small percentage of Medicare adult general medicine patients may be reclassified as inpatients under the new rules, the net effect would be that many more hospital encounters will be billed under observation rather than inpatient status. These findings assume overall length of stay will remain unchanged under these rules, an assumption that may not hold true given the financial losses we predicted hospitals may face under this rule,[12] and potential pressures on individual physicians providing patient care.

Medicare has prioritized auditing <2‐midnight inpatient encounters under the assumption that many short inpatient encounters are actually misclassified observation encounters,[4, 8, 9] prompting us to investigate whether this was the case in our patient population. Although it did not use ICD‐9 diagnosis codes, the OIG report suggested that short‐stay inpatients and observation patients may be clinically similar.[8, 13] Using ICD‐9 codes, we found no overlap between the top ICD‐9 codes for adult general medicine <2‐midnight inpatient and observation encounters, and only 1 of 5 shared codes for these encounters across all service lines. These findings are counter to the OIG report, and suggest that <2‐midnight inpatients are different from observation patients at our institution, and that <2‐midnight inpatients should not be arbitrarily reclassified as observation based solely upon LOS.

We also found that the majority of top ICD‐9 codes within observation were the same regardless of LOS, suggesting that LOS does not reliably differentiate clinically different observation populations that merit different insurance coverage (Medicare Part A for 2‐midnight encounters, Medicare Part B for <2‐midnight encounters). This lack of a clear cut point may drive an overall increase in LOS to achieve 2 midnights, as common diagnosis codes can be justified for both <2‐midnight or 2‐midnight observation stays.

Finally, we found that external factors, such as the time of day and specific day (weekday vs weekend) of hospitalization, impact the likelihood of achieving a 2‐midnight stay. Patients hospitalized earlier in the day were less likely to span 2 midnights compared to later‐day encounters, suggesting that use of a full working day as the day of presentation is harmful to a patient's chance of gaining inpatient status. Observation hospitalizations starting on a weekend day were more likely to achieve 2 midnights, which likely reflects different resource allocation and hospital efficiency on weekends, yet it is unlikely that weekend midnights, if associated with any delay in care, will be counted toward a patient's cumulative 2‐midnight total. The CMS has further indicated that midnights accrued prior to transfer from 1 hospital to another will not count toward a cumulative 2‐midnight stay. Although it would seem likely that patients requiring transfer to a tertiary care center would meet inpatient criteria, 9% of our observation encounters were transferred from other acute care hospitals, and many will have lost attributable midnights accrued prior to transfer to our hospital. Taken together, our findings strongly suggest that issues entirely unrelated to diagnosis or clinical status will impact whether hospitalized patients will be classified as inpatient or observation under the new rule.

This study has several limitations. Our data were limited to a single Midwestern tertiary care academic medical center, and may not be applicable to other healthcare settings. Second, 1.5% (466) of our inpatient stays and 56 (1.2%) of our observation encounters lacked a referral source in our administrative database, although these deficiencies would not likely change the conclusions. Finally, the total number of observation encounters starting before 8:00 _am and staying 2 midnights was small and therefore potentially subject to confounding. However, despite the fact that encounters beginning in the early morning hours may be different in ways unable to be measured in this study, encounter ICD‐9 codes were similar regardless of time of day.

Despite these limitations, our study raises concerns about the impact of CMS's new time‐driven observation rules on hospital practice patterns. We found distinctly different <2‐midnight inpatient ICD‐9 codes as compared to those for observation encounters. Reclassifying <2‐midnight inpatients as observation may inappropriately shift the financial burden of acute hospitalizations to patients who were previously eligible to receive Medicare inpatient benefits. We also demonstrated a decrease in billable inpatient encounters under the new rules, which may be countered by changes in hospital and provider practice to drive more admissions over the 2‐midnight threshold to avoid financial losses. These changes should be defensible under audit given the ambiguities of the rule we found in this study surrounding time of day of admission, weekend and transfer hospitalizations, and the fact that common observation ICD‐9 codes are similar across LOS. Ironically, the unintended consequence of the new rule may be to drive up hospital LOS, reduce efficiency, and increase the overall cost of care.

Our findings suggest that CMS should define observation care per its original intent: as a means to determine if a patient can safely return home after a short period of additional care.[6] Patients whose conditions necessitate an intensity and level of service beyond this narrow scope should be classified as full inpatients irrespective of LOS, whether that LOS is <2 midnights or 2 midnights. Policies that fail to reflect the original intent of observation status are unlikely to achieve the intended outcome of defining 2 distinct patient populations that merit different services and reimbursement.

Observation hospitalization is a growing phenomenon in the United States healthcare system.[1] For Medicare beneficiaries, observation encounters increased 33.6% from 2004 to 2011, with inpatient encounters decreasing by 7.8% over the same period.[2]

Observation length of stay has also increased. Medicare states that observation care should typically last <24 hours, and in only rare and exceptional cases exceed 48 hours. We and others have showed that observation stays, on average, do not meet this definition.[1, 3] At our institution, historically less than half of observation encounters discharge in under 24 hours, and 1 in 6 stay longer than 48 hours.[3]

The Centers for Medicare and Medicaid Services (CMS) issued a rules change effective October 1, 2013, in response to concern about recent increases in the length of time that CMS beneficiaries spend as hospital outpatients receiving observation services.[4] These rules shifted observation determination from clinical criteria, such as InterQual,[5] to a time‐based rule that hinges on a 2‐midnight cut point. Patients staying <2 midnights, with few exceptions, are now observation, and those staying 2 midnights are inpatients. This is important, as patients hospitalized as observation are technically outpatients, not covered by Medicare Part A hospital insurance, and ineligible for skilled nursing facility benefits.[3, 6]

Although challenges with observation status in general are well described,[1, 3, 7] the potential impact of the 2‐midnight rule is not fully known. The purpose of our descriptive study was to examine how the new rules, retrospectively applied to recent encounters, would affect our hospital encounters, with a separate analysis of Medicare encounters and adult general medicine encounters. First, as CMS predicted a net shift from outpatient to inpatient stays[8] under the new rule, we attempted to determine whether this rule would actually reduce observation encounter frequency. Second, as CMS has cited persistently large improper payment rates in short‐stay hospital inpatient claims[4] and intends to audit <2‐midnight inpatient encounters under the assumption that many are misclassified observation stays,[4, 8, 9] we sought to determine if short‐stay inpatient and observation encounters were truly the same. Third, as insurance coverage will change based on the 2‐midnight cut point, we sought to determine whether 2 midnights separated distinct clinical populations within observation status, making the rule logical and fair. Finally, we sought to determine whether external factors, such as time of admission, day of week of admission, and transfer to our institution would impact how patients may be classified under the new rule.

METHODS

RESULTS

Percentage of Observation Encounters Reaching 2 Midnights Based on Time of Hospitalization

Observation encounters starting before 8:00 _am spanned 2 midnights 13.6% of the time, and those encounters starting after 4:00 _pm crossed 2 midnights 31.2% of the time. Two of the 3 top ICD‐9 codes were the same across LOS, with similar findings for the adult general medicine‐only group (Figure 1).

Figure 1

DISCUSSION

Although CMS predicts that more patients will be classified as inpatients under the new rule, we determined the opposite, consistent with a recent report generated by the Office of the Inspector General (OIG) for 2012 Medicare beneficiaries.[8] Our results did not change when we excluded all surgical encounters to account for possible exclusions based on the surgical inpatient‐only list. Although a small percentage of Medicare adult general medicine patients may be reclassified as inpatients under the new rules, the net effect would be that many more hospital encounters will be billed under observation rather than inpatient status. These findings assume overall length of stay will remain unchanged under these rules, an assumption that may not hold true given the financial losses we predicted hospitals may face under this rule,[12] and potential pressures on individual physicians providing patient care.

Medicare has prioritized auditing <2‐midnight inpatient encounters under the assumption that many short inpatient encounters are actually misclassified observation encounters,[4, 8, 9] prompting us to investigate whether this was the case in our patient population. Although it did not use ICD‐9 diagnosis codes, the OIG report suggested that short‐stay inpatients and observation patients may be clinically similar.[8, 13] Using ICD‐9 codes, we found no overlap between the top ICD‐9 codes for adult general medicine <2‐midnight inpatient and observation encounters, and only 1 of 5 shared codes for these encounters across all service lines. These findings are counter to the OIG report, and suggest that <2‐midnight inpatients are different from observation patients at our institution, and that <2‐midnight inpatients should not be arbitrarily reclassified as observation based solely upon LOS.

We also found that the majority of top ICD‐9 codes within observation were the same regardless of LOS, suggesting that LOS does not reliably differentiate clinically different observation populations that merit different insurance coverage (Medicare Part A for 2‐midnight encounters, Medicare Part B for <2‐midnight encounters). This lack of a clear cut point may drive an overall increase in LOS to achieve 2 midnights, as common diagnosis codes can be justified for both <2‐midnight or 2‐midnight observation stays.

Finally, we found that external factors, such as the time of day and specific day (weekday vs weekend) of hospitalization, impact the likelihood of achieving a 2‐midnight stay. Patients hospitalized earlier in the day were less likely to span 2 midnights compared to later‐day encounters, suggesting that use of a full working day as the day of presentation is harmful to a patient's chance of gaining inpatient status. Observation hospitalizations starting on a weekend day were more likely to achieve 2 midnights, which likely reflects different resource allocation and hospital efficiency on weekends, yet it is unlikely that weekend midnights, if associated with any delay in care, will be counted toward a patient's cumulative 2‐midnight total. The CMS has further indicated that midnights accrued prior to transfer from 1 hospital to another will not count toward a cumulative 2‐midnight stay. Although it would seem likely that patients requiring transfer to a tertiary care center would meet inpatient criteria, 9% of our observation encounters were transferred from other acute care hospitals, and many will have lost attributable midnights accrued prior to transfer to our hospital. Taken together, our findings strongly suggest that issues entirely unrelated to diagnosis or clinical status will impact whether hospitalized patients will be classified as inpatient or observation under the new rule.

This study has several limitations. Our data were limited to a single Midwestern tertiary care academic medical center, and may not be applicable to other healthcare settings. Second, 1.5% (466) of our inpatient stays and 56 (1.2%) of our observation encounters lacked a referral source in our administrative database, although these deficiencies would not likely change the conclusions. Finally, the total number of observation encounters starting before 8:00 _am and staying 2 midnights was small and therefore potentially subject to confounding. However, despite the fact that encounters beginning in the early morning hours may be different in ways unable to be measured in this study, encounter ICD‐9 codes were similar regardless of time of day.

Despite these limitations, our study raises concerns about the impact of CMS's new time‐driven observation rules on hospital practice patterns. We found distinctly different <2‐midnight inpatient ICD‐9 codes as compared to those for observation encounters. Reclassifying <2‐midnight inpatients as observation may inappropriately shift the financial burden of acute hospitalizations to patients who were previously eligible to receive Medicare inpatient benefits. We also demonstrated a decrease in billable inpatient encounters under the new rules, which may be countered by changes in hospital and provider practice to drive more admissions over the 2‐midnight threshold to avoid financial losses. These changes should be defensible under audit given the ambiguities of the rule we found in this study surrounding time of day of admission, weekend and transfer hospitalizations, and the fact that common observation ICD‐9 codes are similar across LOS. Ironically, the unintended consequence of the new rule may be to drive up hospital LOS, reduce efficiency, and increase the overall cost of care.

Our findings suggest that CMS should define observation care per its original intent: as a means to determine if a patient can safely return home after a short period of additional care.[6] Patients whose conditions necessitate an intensity and level of service beyond this narrow scope should be classified as full inpatients irrespective of LOS, whether that LOS is <2 midnights or 2 midnights. Policies that fail to reflect the original intent of observation status are unlikely to achieve the intended outcome of defining 2 distinct patient populations that merit different services and reimbursement.