diff --git a/icu4c/source/i18n/cpdtrans.cpp b/icu4c/source/i18n/cpdtrans.cpp
index 6731779e91a..f3efc084ecd 100644
--- a/icu4c/source/i18n/cpdtrans.cpp
+++ b/icu4c/source/i18n/cpdtrans.cpp
@@ -30,6 +30,8 @@ static const UChar COLON_COLON[] = {0x3A, 0x3A, 0}; //"::"
 
 U_NAMESPACE_BEGIN
 
+const UChar CompoundTransliterator::PASS_STRING[] = { 0x0025, 0x0050, 0x0061, 0x0073, 0x0073, 0 }; // "%Pass"
+
 UOBJECT_DEFINE_RTTI_IMPLEMENTATION(CompoundTransliterator)
 
 /**
@@ -373,16 +375,16 @@ UnicodeString& CompoundTransliterator::toRules(UnicodeString& rulesSource,
         // ::BEGIN/::END blocks) are given IDs that begin with
         // "%Pass": use toRules() to write all the rules to the output
         // (and insert "::Null;" if we have two in a row)
-        if (trans[i]->getID().startsWith("%Pass")) {
+        if (trans[i]->getID().startsWith(PASS_STRING)) {
             trans[i]->toRules(rule, escapeUnprintable);
-            if (numAnonymousRBTs > 1 && i > 0 && trans[i - 1]->getID().startsWith("%Pass"))
-                rule = "::Null;" + rule;
+            if (numAnonymousRBTs > 1 && i > 0 && trans[i - 1]->getID().startsWith(PASS_STRING))
+                rule = UNICODE_STRING_SIMPLE("::Null;") + rule;
 
         // we also use toRules() on CompoundTransliterators (which we
         // check for by looking for a semicolon in the ID)-- this gets
         // the list of their child transliterators output in the right
         // format
-        } else if (trans[i]->getID().indexOf(';') >= 0) {
+        } else if (trans[i]->getID().indexOf(ID_DELIM) >= 0) {
             trans[i]->toRules(rule, escapeUnprintable);
 
         // for everything else, use Transliterator::toRules()
diff --git a/icu4c/source/i18n/cpdtrans.h b/icu4c/source/i18n/cpdtrans.h
index 72e85e1de76..e4688fead5a 100644
--- a/icu4c/source/i18n/cpdtrans.h
+++ b/icu4c/source/i18n/cpdtrans.h
@@ -191,6 +191,9 @@ public:
      */
     static UClassID U_EXPORT2 getStaticClassID();
 
+    /* @internal */
+    static const UChar PASS_STRING[];
+
 private:
 
     friend class Transliterator;
diff --git a/icu4c/source/i18n/translit.cpp b/icu4c/source/i18n/translit.cpp
index d7e7b173c96..3469fe6feea 100644
--- a/icu4c/source/i18n/translit.cpp
+++ b/icu4c/source/i18n/translit.cpp
@@ -929,7 +929,7 @@ Transliterator::createInstance(const UnicodeString& ID,
     U_ASSERT(list.size() > 0);
     Transliterator* t = NULL;
     
-    if (list.size() > 1 || canonID.indexOf(";") >= 0) {
+    if (list.size() > 1 || canonID.indexOf(ID_DELIM) >= 0) {
         // [NOTE: If it's a compoundID, we instantiate a CompoundTransliterator even if it only
         // has one child transliterator.  This is so that toRules() will return the right thing
         // (without any inactive ID), but our main ID still comes out correct.  That is, if we
@@ -1063,7 +1063,7 @@ Transliterator::createFromRules(const UnicodeString& ID,
         if (parser.compoundFilter != NULL) {
             UnicodeString filterPattern;
             parser.compoundFilter->toPattern(filterPattern, FALSE);
-            t = createInstance(filterPattern + ";"
+            t = createInstance(filterPattern + UnicodeString(ID_DELIM)
                     + *((UnicodeString*)parser.idBlockVector->elementAt(0)), UTRANS_FORWARD, parseError, status);
         }
         else
@@ -1095,7 +1095,8 @@ Transliterator::createFromRules(const UnicodeString& ID,
             }
             if (!parser.dataVector->isEmpty()) {
                 TransliterationRuleData* data = (TransliterationRuleData*)parser.dataVector->orphanElementAt(0);
-                transliterators.addElement(new RuleBasedTransliterator((UnicodeString)"%Pass" + (passNumber++),
+                transliterators.addElement(
+                    new RuleBasedTransliterator(UnicodeString(CompoundTransliterator::PASS_STRING) + (passNumber++),
                     data, TRUE), status);
             }
         }
diff --git a/icu4c/source/i18n/transreg.cpp b/icu4c/source/i18n/transreg.cpp
index fc94631a390..da6356f97c3 100644
--- a/icu4c/source/i18n/transreg.cpp
+++ b/icu4c/source/i18n/transreg.cpp
@@ -1223,7 +1223,7 @@ Transliterator* TransliteratorRegistry::instantiateEntry(const UnicodeString& ID
             UVector* rbts = new UVector(status);
             int32_t passNumber = 1;
             for (int32_t i = 0; U_SUCCESS(status) && i < entry->u.dataVector->size(); i++) {
-                Transliterator* t = new RuleBasedTransliterator((UnicodeString)"%Pass" + (passNumber++),
+                Transliterator* t = new RuleBasedTransliterator(UnicodeString(CompoundTransliterator::PASS_STRING) + (passNumber++),
                     (TransliterationRuleData*)(entry->u.dataVector->elementAt(i)), FALSE);
                 if (t == 0)
                     status = U_MEMORY_ALLOCATION_ERROR;